The study was approved by the Ethics Committee of Centro Hospitalar Universitário de Lisboa Central and was conducted according to the Declaration of Helsinki. All participants signed written informed consent. This was a single-center, prospective observational study. Consecutive patients with HFrEF that underwent remote monitoring based on CardioMEMS™ system between January 2020 and December 2022 were included.

The eligibility criteria for CardioMEMS™ system implantation were the diagnosis of HF, NYHA class ≥II despite optimal medical therapy according to the published guidelines,17 at least one HFH in the previous year and good patient compliance as judged by HF team.

Exclusion criteria included a history of recurrent pulmonary embolism or deep vein thrombosis, inability to tolerate RHC, major cardiovascular events (including myocardial infarction and stroke) within the previous three months, cardiac resynchronization therapy (CRT) implanted within the previous three months, glomerular filtration rate <25 mL/min per 1.73 m2 (measured within two weeks of pressure sensor implant), congenital heart disease or mechanical right heart valve, inclusion on the waiting list for heart transplantation, need for ventricular assist device, active infection and known coagulation disorders.

Baseline clinical, demographic and laboratory data were collected at the study inclusion, as well as invasive hemodynamic data from the RHC procedure performed during implantation of the CardioMEMS™ sensor.

According to the protocol, a transthoracic echocardiography (TTE) study (as detailed below) was performed at three, six, 12, 18 and 24 months after the CardioMEMS™ system implantation.

Patients performed daily readings using the CardioMEMS™ system, as described below. Invasive hemodynamic data were obtained, including sPAP, dPAP, mPAP, heart rate (HR) and estimated CO. For this study, additional readings were performed immediately before each TTE study, and such data were used for the analysis.

A routine protocol from the echocardiographic laboratory at our center for conventional M-mode, two-dimensional, Doppler, and tissue Doppler measurement was used.

LV diameters were measured from the parasternal long-axis view. LV end-systolic and end-diastolic volumes were measured and indexed to body surface area (BSA). LVEF was calculated using the Simpson biplane method. Left atrium (LA) volume was measured by a biplane area-length method from the apical four and two-chamber views and indexed to the BSA. Mitral E and A peak velocity were measured and the ratio of early diastolic LV inflow velocity to atrial-systolic velocity (E/A) was calculated. The average tissue Doppler-derived early diastolic mitral annular velocity (e′) was obtained from the mitral annulus’ septal and lateral sides. The average ratio of early diastolic LV inflow velocity to early diastolic mitral annular velocity (E/e′) was calculated. Tricuspid annular plane systolic excursion (TAPSE) was measured by M-mode and sPAP was estimated based on the tricuspid regurgitant jet velocity.

Left ventricular global longitudinal strain analysis was performed with two-dimensional speckle tracking-echocardiography with imaging acquisition at 50–70 frames/s. LV endocardial border was manually traced at the end-diastole frame (first frame of mitral valve coaptation) from three apical views (four-, two- and three-chamber views). To ensure accurate tracking, the TTE operator assessed the detected region of interest (ROI) and manually modified the ROI as appropriate. LVGLS was calculated by averaging all available segments at the time of peak systolic strain.

Briefly, the CardioMEMS™ system consists of three main components: an implantable PA sensor, a patient and hospital electronic system, and a digital database. The PA sensor is a small device placed permanently in a PA branch and features nitinol loops for secure positioning. It provides hemodynamic data, including PAP waveform, sPAP, dPAP, mPAP, HR and CO estimation.

The CardioMEMS™ sensor was implanted in the PA according to manufacturer instructions16 and calibrated with a RHC using a balloon wedge-pressure catheter and indirect Fick CO calculation. The CardioMEMS™ system measures PAP, which is highly correlated with PAP obtained using a Swan-Ganz catheter.18 The CO was estimated using a proprietary algorithm (Abbott Scientific, Atlanta, GA, USA) based on the PAP waveform, mPAP, HR, and a reference CO measured at the implant.19 Even though estimated CO is not currently approved for regular clinical use, the algorithm had undergone prior development, refinement, and retrospective testing using clinical RHC data. These tests showed that the estimated CO is non-inferior to currently employed CO measurement methods.20

Categorical variables are presented as frequency and percentages, and continuous variables as means and standard deviations or medians and interquartile ranges, as appropriate. Normal distribution was checked using the Shapiro–Wilk test.

We performed uni- and multivariate analyses to assess whether LVGLS was associated with hemodynamic parameters (sPAP, dPAP, mPAP and CO), irrespective of other echocardiographic parameters. First, we performed a univariate analysis using Spearman's correlation to assess which echocardiographic parameters, including LVGLS, were associated with each of the hemodynamic parameters. We then performed a multivariable analysis (backward stepwise technique) for the hemodynamic parameters that had a significant correlation with LVGLS in the previous analysis, adjusting for other echocardiographic covariates. Variables with a p value <0.05 in the univariate analyses were included in the multivariable model. For covariate adjustment, the following echocardiographic parameters were considered: LV end-diastolic diameter (LVEDD), LV end-systolic diameter (LVESD), indexed LV end-diastolic volume (LVEDV), indexed LV end-systolic volume (LVESV), interventricular septum (IVS) thickness, LA volume, LVEF, TAPSE, sPAP, mean E/e′ ratio, E-wave velocity, A-wave velocity and E/A ratio.

In addition, after the multivariate regression models had been performed, we computed a post hoc power analysis for each model obtained. We used the G*Power 3.1.9 software for this purpose. We used a F test for linear multiple regression fixed model, R2 deviation from zero to perform a post hoc power analysis, assuming a medium effect size f2 of 0.15, a type α error probability of 0.05, a sample size of 46 and the number of predictors according to each final model. The power (1−β error probability) was then calculated for each model.

Finally, an exploratory analysis was carried out to assess the potential association between the variation in left ventricular global longitudinal strain (LVLGS) and variation in the hemodynamic parameters (sPAP, dPAP, mPAP and CO) during follow-up. The variation in LVGLS and hemodynamic parameters was calculated as the absolute difference between two consecutive measurements for each patient. The Spearman's correlation was used to test the correlations between the mean variation in LVGLS between consecutive assessments and the mean variation (corresponding to the same follow-up period) in hemodynamic parameters.

IBM SPSS Statistics® (version 26) was used for statistical analyses. A p value <0.05 was considered significant.

A total of 12 patients were included in the study and a total of 46 TTE were performed during a median follow-up time of 20.5 months. Table 1 summarizes the baseline characteristics of the sample. The mean age was 67±11 years, 11 (92%) were male and 8 (67%) had ischemic dilated cardiomyopathy. Mean LVEF was 28.5±8%, mean LVGLS was −7.3±3% and mean NT-proBNP was 5032±4518.1 pq/mL. Five (42%) patients had past medical history of atrial fibrillation (AF), but only four out of 46 TTE studies were performed in AF rhythm. All patients were on optimized guideline-directed medical therapy. Nine (75%) patients had an implantable cardiac device (ICD) and 2 (16.7%) had CRT. Five patients had more than one HFH and 4 (33%) had received a levosimendan infusion in the previous year.

At the time of CardioMEMS™ implantation, mean dPAP was 19±9 mmHg, mean sPAP was 46±24 mmHg, mean mPAP was 28±18 mmHg, mean PCWP was 21±9 mmHg and, mean HR was 65±10 beats per minute and mean CO was 4.8±1.3 L/min (Table 2).

A univariate analysis was performed to assess the correlation between echocardiographic parameters, including LVGLS, and invasive hemodynamic parameters obtained using the CardioMEMS™ (Table 3).

Left ventricular global longitudinal strain was significantly correlated with both dPAP (r=0.40, p=0.04) and CO (r=−0.43, p=0.03), but neither with sPAP nor with mPAP. Considering such results, the echocardiographic variables correlated with dPAP and/or CO, other than LVGLS, are presented.

LVESD (r=0.39, p=0.03), LVEDV (r=0.41, p=0.007), LVESV (r=0.31, p=0.04), LA volume (r=0.39, p=0.01), echocardiographic-derived sPAP (r=0.41, p=0.01), peak E-wave velocity (r=0.37, p=0.016) and E/A ratio (r=0.65, p<0.001) were significantly correlated with dPAP. Only LVESV showed a significant correlation with CO (r=−0.35, p=0.02), in addition to LVGLS.

Considering the results obtained from the univariate analysis, multiple linear regression models were performed for dPAP and CO as the dependent variables (Table 4).

LVGLS and echocardiographic-derived sPAP were the two independent predictors of invasive dPAP. This model explains 60% of dPAP value (R2=0.60, adjusted R2=0.56, p<0.001). A 1% higher LVGLS absolute value is associated with a 1.3 mmHg higher dPAP (p<0.001), when adjusted for echocardiographic-derived sPAP. LVGLS was the only independent predictor of CO. The model explains 36% of CO value (R2=0.362, p=0.001). A 1% higher LVGLS absolute value is associated with a lower 0.38 L/min CO.

Considering the two final models mentioned above, the statistical power analysis of the dPAP regression model was determined to be 0.62 and for the CO regression model was 0.73.

In the exploratory analysis, the variation in LVGLS was significantly correlated with a corresponding variation with invasive dPAP (r=0.60, p=0.017) (Table 5).