Background: Cardiac transthyretin (ATTR) amyloidosis is an often underdiagnosed and potentially fatal disorder associated with poor survival. The National Amyloidosis Centre (NAC) staging system, based on NT-proBNP level and eGFR value, discriminates patients according to survival rates. However, NAC stage II involves a heterogenous group of patients with variable prognosis. This retrospective single-center study was set up to explore the potential role of myocardial work (MW) analysis to enhance risk stratification of ATTR patients prior to therapy. Methods and Results: 37 patients diagnosed with ATTR between March 2021 and August 2023 were included. Baseline NT-proBNP and eGFR values were collected and LVEF, GLS and MW parameters were obtained from stored echocardiographic images. Patients were categorized per NAC stage (16 NAC I, 13 NAC II and 8 NAC III). Whereas the survival rate in NAC II and NAC III was significantly worse than in NAC I (p = 0.031 and p = 0.045 respectively), no significant difference was found between NAC II and III. In the ROC analysis, GCW proved to be the best survival predictor (AUC: 0.7) with optimal cut-off value 1294 mmHg%. Patients from NAC stage II were re-stratified according to GCW cut-off into HIGH RISK together with patients from NAC III or LOW RISK together with patients from NAC I. Patients in the HIGH RISK group exhibited a significantly worse prognosis with only 40 % survival at 2 years follow-up. Conclusion: Our results demonstrate the advantages of incorporating MW analysis, particularly the use of a GCW cut-off, in the baseline risk stratification of ATTR patients.

Background Transcatheter aortic valve replacement(TAVR) has shown clear survival benefits in severe aortic valve stenosis(AS). However, patients unable to recover left ventricle function remain at risk with poor long-term survival. This single-center prospective study aims to analyze the supplementary benefits of myocardial work(MW) assessment for baseline risk stratification in patients with severe AS referred for TAVR. Methods A total of 110 patients with severe AS referred for TAVR were included in the study. Baseline ECG data, transthoracic echocardiographic(TTE) images and blood samples were obtained. The TTE examination was repeated one day and one month after valve replacement. The primary outcome of the study was a composite endpoint consisting of all-cause mortality and HF hospitalization. Results During a mean follow-up period of 521 ± 343 days, 29patients(26.4 %) reached the composite endpoint. Baseline troponins, NT-proBNP, sST2, GWI and GCW showed statistically significant differences between groups. Patients with a baseline GWI<2323 mmHg% (sensitivity 0.63 and specificity 0.76)had significantly worse outcome following TAVR. A basic predictive model included QRS-length, TAPSE, LAVI and E/e’. The addition of biomarkers did not yield any further advantages whereas incorporating the GWI cut-off value of 2323 mmHg% significantly enhanced the predictive value. Although there were no significant changes in LVEF and GLS, all patients exhibited a significant reduction in GWI and GCW immediately after TAVR. Conclusion Our findings provide evidence for the enhanced usefulness of MW analysis in the initial risk stratification of patients with severe AS referred for TAVR. Specifically, a baseline GWI<2323 mmHg% demonstrates an independent predictor associated with increased incidence of all-cause mortality and HF hospitalization following TAVR.

Abstract Background Heart failure (HF) frequently causes an imbalance in heart and kidney functions, and the prognosis of heart and kidney disease can worsen one another. In addition to the prognostic role of chronic kidney disease (CKD), dynamic changes in renal function have been recognized to portend a poor prognosis in HF patients. Purpose This study aimed to investigate the prevalence of impaired renal function at HF diagnosis and its association with short and long-term outcomes in a real-world cohort of HF patients. In addition, we assessed how dynamic changes in renal function after HF diagnosis might predict HF prognosis. Methods Using a natural language processing algorithm, clinical data of patients with at least one diagnosis of HF (index event) were retrospectively collected. Patients for which the most recent estimated glomerular filtration rate (eGFR) value prior to the index date was ≤ 60 mL/min/1.73m2 and/or had a documented medical history of CKD (including dialysis) were considered as having CKD. A linear mixed effects model was used to show the longitudinal eGFR trajectories during the follow-up period (after index). A Cox proportional hazard model was used to analyse the association between baseline variables and mortality. Finally, joint modelling was used to combine the linear mixed eGFR sub-model with the survival sub-model, to estimate the effect of the current eGFR value and the eGFR slope of the year prior on mortality. Results The study population consisted of 1992 patients with a mean (±SD) age of 74.8 ± 11.7 years and 58% males. In total, 17% had HF with preserved ejection fraction (HFpEF, n = 339) and 70% (n = 1156) presented CKD at the index event. The median follow-up time was 3.16 years and 13% (n = 198) of the patients were deceased after two years. The group of 2-year survivors had significantly lower proportion of CKD at index (55 vs 76%, p < 0.001). The calculated mean yearly decrease of the eGFR value was 4.2 mL/min/1.73m². The multivariate Cox PH model identified age, CVA, diastolic blood pressure, HDL, and log10proBNP as baseline predictors of mortality. Joint modelling showed that a 10 mL/min/1.73m2 lower current eGFR value amongst patients was associated with an increased mortality hazard of 1.30 ± 0.06 (p < 0.001). Additionally, a decrease of 10 mL/min/1.73m2 in eGFR during the year prior indicated an increased mortality hazard of 1.87 ± 0.19 (p < 0.001). Conclusion In a real-world cohort of HF patients, the prevalence of CKD is high and CKD is an independent predictor of mortality. The current eGFR value and the year prior’s eGFR slope adds additional prognostic information and helps to better fine tune the individual mortality hazard. Based on the findings of this analysis, patients with worse prognosis might benefit from a closer and more vigilant follow-up.

Serial transthoracic echocardiographic (TTE) assessment of LVEF and GLS are the gold standard in screening Cancer Therapeutics-Related Cardiac Dysfunction (CTRCD). Non-invasive left-ventricle (LV) pressure–strain loop (PSL) emerged as a novel method to quantify Myocardial Work (MW). This study aims to describe the temporal changes and longitudinal trajectories of MW indices during cardiotoxic treatment. We included 50 breast cancer patients with normal LV function referred for anthracycline therapy w/wo Trastuzumab. Medical therapy, clinical and echocardiographic data were recorded before and 3, 6, and 12 months after initiation of the chemotherapy. MW indices were calculated through PSL analysis. According to ESC guidelines, mild and moderated CTRCD was detected in 10 and 9 patients, respectively (20% CTRCDmild, 18% CTRCDmod), while 31 patients remained free of CTRCD (62% CTRCDneg). Prior to chemotherapy MWI, MWE and CW were significantly lower in CTRCDmod than in CTRCDneg and CTRCDmild. Overt cardiac dysfunction in CTRCDmod at 6 months was accompanied by significant worse values in MWI, MWE and WW compared to CTRCDneg and CTRCDmild. MW features such as low baseline CW, especially when associated with a rise in WW at follow-up, may identify patients at risk for CTRCD. Additional studies are needed to explore the role of MW in CRTCD.

Abstract Aim This prospective longitudinal study analyses the potential role of Myocardial Work in early detection of cardiotoxicity during chemotherapy and its added value for prognosis and patients' risk stratification. Methods We enrolled 47 consecutive female patients with HER2-positive breast cancer referred for anti-cancer therapy based on anthracycline and taxane. Patients with depressed LV function at baseline were excluded. Medical therapy, clinical parameters and echocardiographic data were recorded at baseline and at 3, 6, 12 months follow-up. Additionally, cuff blood pressure was measured at the time of 2D-TTE examination and adequate echocardiographic images were stored for off-line analysis. Results CTRCD was detected in 17 patients (36%) while 30 patients remained free of CTRCD (64%). There were no intergroup differences for age, body mass index, resting heart rate and brachial arterial pressure. Both groups presented unaltered LV systolic function after 3 months follow-up yet overt cardiac dysfunction showed up in the CTRCD group at 6 months with significant decline in LVEF, GLS, MWI, MWE and CW from baseline values (LVEF, %: 56.0±4.1 vs 52.2±6.5; GLS, %: −20.9±1.9 vs −17.6±3.2; MWI, mmHg%: 2125±348 vs 1704±620; MWE, %: 95±2.6 vs 93±3.9 and CW, mmHg%: 2562±3567 vs 2212±455, p<0.05). Additionally, GLS, MWI and MWE at 6 months were significantly worse in the CTRCD group vs non-CTRCD group (GLS, %: −17.6±3.2 vs −20.6±1.8; MWI, mmHg%: 1704±620 vs 2087±347; MWE, %: 93±3.9 vs 96±1.5, p<0.05). Depressed LV systolic function persisted after 1 year follow-up (Figure 1). After 3 months, only de relative change in GLS and WW from baseline were significantly worse in CTRCD vs non-CTRCD (ΔGLS: +3.7±11 vs −3.9±10, ΔWW: +46.1±83 vs +2.2±45). Whereas no correlation was found, the combination of both ΔGLS and ΔWW at 3 months showed stronger prognostic value for CTRCD than each parameter alone, AUC of 0.72 (Figure 2). Conclusion These findings point the superiority of Myocardial Work for early type 1 CTRCD detection in comparison to the current diagnostic tools. Additionally, we suggest the add-on value of ΔWW on top of ΔGLS quantification for better patient risk stratification. These are promising results for better clinical surveillance of cardiac function during cancer treatment. Funding Acknowledgement Type of funding sources: None.

Risk stratification in patients with a new onset or worsened heart failure (HF) is essential for clinical decision making. We have utilized a novel approach to enrich patient level prognostication using longitudinally gathered data to develop ML-based algorithms predicting all-cause 30, 90, 180, 360, and 720 day mortality. In a cohort of 2449 HF patients hospitalized between 1 January 2011 and 31 December 2017, we utilized 422 parameters derived from 151 451 patient exams. They included clinical phenotyping, ECG, laboratory, echocardiography, catheterization data or percutaneous and surgical interventions reflecting the standard of care as captured in individual electronic records. The development of predictive models consisted of 101 iterations of repeated random subsampling splits into balanced training and validation sets. ML models yielded area under the receiver operating characteristic curve (AUC-ROC) performance ranging from 0.83 to 0.89 on the outcome-balanced validation set in predicting all-cause mortality at aforementioned time-limits. The 1 year mortality prediction model recorded an AUC of 0.85. We observed stable model performance across all HF phenotypes: HFpEF 0.83 AUC, HFmrEF 0.85 AUC, and HFrEF 0.86 AUC, respectively. Model performance improved when utilizing data from more hospital contacts compared with only data collected at baseline. Our findings present a novel, patient-level, comprehensive ML-based algorithm for predicting all-cause mortality in new or worsened heart failure. Its robust performance across phenotypes throughout the longitudinal patient follow-up suggests its potential in point-of-care clinical risk stratification.

A membrane-aerated biofilm reactor (MABR) capable of simultaneous nitrification and denitrification in a single reactor vessel was developed to investigate the characteristics of nitrogen removal from high-strength nitrogenous wastewater, and biofilm analysis using microelectrodes and the fluorescence in situ hybridization (FISH) technique was performed. Mean removal percentages of total organic carbon (TOC) and nitrogen were 96% and 83% at removal rates of 5.76 g-C m−2 d−1 and 4.48 g-N m−2 d−1, respectively. For stable removal efficiency, constant washing of the biofilm was needed. Dissolved oxygen microelectrode measurement revealed that the biofilm thickness was about 1600 μm, and that oxygen penetrated about 300 to 700 μm, from the outer surface of the membrane. Furthermore, FISH analysis revealed that ammonia-oxidizing bacteria (AOB) were located near the outer surface of the membrane, whereas other bacteria were located from the inner to the outer part of the biofilm. Combining these results demonstrated that simultaneous nitrification and denitrification occurred in the biofilm of the MABR system. In addition, stoichiometric analysis revealed that after 130 d−1, the free ammonia (FA) concentration ranged within the concentration causing inhibition of the growth of nitrite oxidizing bacteria (NOB) and that AOB consumed 86% of the oxygen supplied through the intra-membrane. These results indicate that nitrogen removal not via nitrate but via nitrite was mainly achieved in the MABR system.

Background. A pulmonary capillary wedge pressure (PCWP) >18 mm Hg following volume load has been proposed as a partition value for the detection of heart failure with preserved ejection fraction. As hemodynamic changes in filling pressures (FP) have been attributed to a nitric oxide (NO)-mediated rightward shift of the pressure-volume relationship, we investigated the hemodynamic response to volume load in heart transplant recipients (HTx) and examined the role of inducible NO synthase (iNOS) gene expression on diastolic function changes. Methods. In 36 HTx, FPs were measured before and after volume load, following which Starling curves were constructed using PCWP and cardiac index (CI). Patients were categorized into those with normal (group A, n = 21) and abnormal hemodynamics (group B, n = 15, PCWP >15 mm Hg at rest or >18 mm Hg following volume load). For the establishment of the potential role of NO, endomyocardial iNOS gene expression level was measured. Results. Except for PCWP (P < 0.001) and mean pulmonary artery pressure (P < 0.001) no differences in age, baseline characteristics, and ejection fraction were observed between both groups, and volume load significantly increased PCWP in both groups (group A: P < 0.001 and group B: P < 0.001) without any change in heart rate. Interestingly, volume load significantly increased CI in group A (P < 0.001) but not in group B (P = 0.654), and the Starling curves revealed a higher CI at any given PCWP in group A together with significantly higher iNOS gene expression (P = 0.009). Conclusions. In HTx, volume load increases FP and unmasks the presence of left ventricular diastolic dysfunction. Interestingly, following saline load group B shows a blunted Starling response, with higher PCWP and lack of CI increase at any given PCWP. The higher iNOS gene expression level in group A suggests a potential role of NO as mediator of diastolic function.

Abstract Introduction Abnormal GLS values as well as high plasma levels of NT-proBNP previous to TAVR are independent predictors for higher peri-procedural mortality. Moreover, in a subgroup of TAVR patients LV function does not recover following the procedure. Until today, it is still unclear how to predict impaired post-procedural LV function for optimal clinical patient's management. Purpose This study was set up to assess the predictive value of baseline GLS and NT-proBNP levels on LV function recovery (LVfr) in a cohort of patients with severe AS referred for TAVR. Methods A total of 25 patients (9 male, 84±5 yo, EF 50±11%) with severe AS (AVA 0.6±0.3 cm2, MPG 49±16 mmHg) referred for TAVR were included. Blood analysis and TTE were performed before intervention (baseline, bl) and at follow-up (fu). Myocardial work was analysed offline integrating the longitudinal strain and afterload pressure (SBP + AVPmean). LVfr was defined as GLS <−19% at fu. The median values at bl of NT-proBNP (1781 ng/L) and GLS (−15%) were taken as cut-off to categorize patients in 4 groups: NT-proBNPhighGLShigh, NT-proBNPlowGLShigh, NT-proBNPhighGLSlow and NT-proBNPlowGLSlow. The ROC curve analysis for prediction of LVfr after TAVR were performed. Results LV function recovered in 13 patients (52%). Despite similar EF and global MWI after TAVR, the LV contraction became more efficient as evidenced by a significant improvement (bl vs fu, p<0.05) in GLS (−14±4.5 vs −18±4.2%), MWW (400±510 vs 157±107 mmHg%) MWE (88±6 vs 92±6%) together with a reduction in afterload pressure (203±38 vs 156±22 mmHg, p<0.05). In the NT-proBNPlow groups, GLS (−15±4 vs −20±3%, p<0.05) and MCW (2166±874 vs 2978±634 mmHg%, p<0.05) at fu were significantly better when compared to the NT-proBNPhigh groups. Likewise, the GLSlow groups showed higher EF (47±10 vs 54±6%, p<0.05) and MCW (2181±832 vs 2961±715 mmHg%, p<0.05) than the GLShigh groups at fu. Interestingly, the GLSlow groups had lower LVESV (57±38 vs 29±10 ml, p<0.05) and LVEDV (113±49 vs 80±20 ml, p<0.05) post-TAVR than the GLShigh groups which suggests a positive remodelling following afterload reduction. At the ROC curve analysis, combined GLS and NT-proBNP at bl were better predictors for LVfr than each parameter alone, AUC 0.86 (Fig. 1). Additionally, only 20% LVfr was seen in the NT-proBNPhighGLShigh group in contrast to 67–75% in the other groups. Conclusion Elevated afterload in severe AS leads to a physiological reduction of GLS. Although the decrease in afterload after TAVR beneficially affects GLS and may lead to LVfr, this was not observed in a subgroup of patients with high NT-proBNP levels in whom GLS remained impaired at follow-up. We speculate that myocardial tissue damage and fibrosis due to long lasting high pressure exposure may partly be responsible for this observation. The combination of pre-procedural NT-proBNP levels and GLS shows strong predictive potential for LVfr after TAVR and larger studies are warranted for further evaluation and cut-off values determination. Funding Acknowledgement Type of funding sources: Foundation. Main funding source(s): Cardiovascular Research Center Aalst (npo)

Abstract Introduction Breast cancer patients receiving anthracyclines are particularly prone to develop cancer therapeutics-related cardiac dysfunction. Early detection of cardiotoxicity onset is required for optimal timing of cardio protection treatment. The latest guidelines consider a relative reduction of 15% in global longitudinal strain (GLS) from baseline as risk for cardiotoxicity. Nevertheless, the more recent Myocardial Work Index (MWI) offers a load-independent tool for detection of subclinical heart failure (HF). However, data in cancer patients are still scarce. Purpose This study analyses the predictive value of MWI for cardiotoxicity diagnosis after 6 months chemotherapy. Methods The study population consists of breast cancer patients referred for chemotherapy with anthracyclines and taxanes. Patients with a history of HF previous to chemotherapy or depressed LV function at baseline were excluded. Echocardiography was performed before onset of the chemotherapy (baseline) and after 6 months follow-up. LVEF, GLS and MWI were assessed offline using EchoPAC software. The values at baseline and 6 months follow-up were pairwise compared to detect subclinical cardiac dysfunction. LVEF, GLS and MWI means at baseline were taken as cut-off to compare the predictive value of each parameter. Moreover, patients were categorized in one group with GLS reduction >15% (Group 1) and one group with GLS reduction <15% (Group 2). Results From April 2016 to July 2020, 28 women with breast cancer were included (age 54±11 years, LVEF 58±4%, GLS −21±2%, MWI 2160±308 mmHg). All patients underwent the same standard chemotherapy protocol (4xEC, 12xTaxol). No difference in baseline characteristics between group 1 (n=13) and group 2 (n=15) was observed. At 6 months follow up a significant decrease in LVEF (53±8%, p=0.003), GLS (−19±3%, p=0.002) and MWI (1920±391 mmHg, p=0.005) was shown without any change in blood pressure. However, while mean LVEF and GLS at baseline did not predict any significant change, patients with MWI under the mean value at baseline (n=15) presented significant lower LVEF (50±8 vs 57±6% p=0.006), GLS (−17±3 vs −20±2%, p=0.01), MWI (1733±320 vs 2136±362 mmHg, p=0.005) after 6 months. Additionally, both groups had similar MWI at baseline (2148±335 mmHg vs 2170±294 mmHg, p=0.85), whereas those patients with GLS reduction >15% showed significant lower MWI after 6 months (1694±332 mmHg vs 2116±334 mmHg, p=0.003, Figure 1). Conclusions At 6 months follow up, a decline of the LV systolic function as side effect of chemotherapy can be seen. MWI at baseline shows the best predictive value for development of cardiotoxicity, in comparison to LVEF and GLS. Further studies are warranted to better understand the role of MWI for early detection of cardiotoxicity and its clinical relevance. Funding Acknowledgement Type of funding sources: Private hospital(s). Main funding source(s): Onze-Lieve-Vrouw hospital in Aalst (Belgium)

Abstract Aims We investigated the prognostic relevance of serpin peptidase inhibitor, clade A member 3 (SERPINA3) in patients admitted with a de novo or worsened heart failure (HF). Methods and results In the first stage, 83 HF‐related left ventricular (LV) transcripts were examined in patients with congestive cardiomyopathy (CCMP, n = 44) who died within 5 years and compared with age‐matched and haemodynamically matched CCMP survivors (n = 39) and controls with normal LV function (n = 17). Among 14 differentially expressed transcripts, myocardial gene and circulating SERPINA3 levels were up‐regulated in non‐survivors vs. survivors (2.40 ± 3.66 vs. 0.36 ± 0.22 units, P < 0.01 and 334.7 ± 138.7 vs. 228.2 ± 83.1 μg/mL, P < 0.01, respectively). While no significant transmyocardial gradient was detected, cytokine stimulation of human endothelial cells induced SERPINA3 secretion. In an independent validation cohort with a de novo or worsened HF (n = 387), circulating SERPINA3 levels > 316 μg/mL were associated with increased all‐cause mortality {hazard ratio [HR] [95% confidence interval (CI)]: 2.4 [1.5–3.9], P = 0.0002} and its composite with unplanned cardiovascular readmission [HR (95% CI): 2.0 (1.2–3.3), P = 0.004]. Patients with elevated SERPINA3 levels and elevated either N‐terminal pro brain natriuretic peptide or ST2 showed worse freedom from both endpoints. In a multivariate analysis, including established clinical risk factors, SERPINA3 remained independent predictor of all‐cause mortality together with age, gender, ST2, glomerular filtration, and pulmonary capillary wedge pressure. Conclusion In patients with a de novo or worsened HF, increased SERPINA3 levels > 316 μg/mL are associated with increased mortality or unplanned cardiac readmission. Elevated SERPINA3 levels on top of established clinical predictors appear to identify a subgroup of HF patients at higher mortality risk. Prospective studies should further validate its value in prognostic stratification of HF.

A 53-year-old heart transplant recipient with a past medical history of surgically treated acute type A dissection of the donor aorta 10 years following heart transplantation, presented with a Card...

Abstract Background Heart failure (HF) is a heterogenous syndrome with complex pathophysiology. Biomarkers and clinical risk scores often fail to provide optimal patient-level precision in the prognostic stratification. As utilizing single observational timepoint, they do not capture the entire care pathway with variations in individual patient management. Electronic patient records provide an opportunity to develop new artificial intelligence (AI) strategies for comprehensive prognostic re-stratification reflecting diagnostic and therapeutic management. Purpose We sought to use deep artificial intelligence (AI) and develop an unbiased predictive algorithm for all-cause mortality in a cohort of patients hospitalized with a de novo or worsened HF. Methods In a cohort of 2449 HF patients hospitalized between 2011–2017, we utilized 151 451 patient exams from 422 parameters. They included clinical phenotyping, medication, ECG, laboratory, echocardiography, catheterization data or percutaneous and surgical interventions gathered on a routine clinical basis reflecting standard of care as captured in individual electronic records. The AI model development consisted of 101 iterations of repeated random subsampling splits into balanced training and validation sets. Results AI models yielded performance ranging from 0.83 to 0.89 AUC on the outcome-balanced validation set in predicting all-cause mortality at 30-, 90-, 180-, 360- and 720-day time-limits (Figure 1). The primary endpoint, 1-year mortality prediction model, recorded an 0.85 AUC accuracy. We observed stable model performance across all HF phenotypes: HFpEF 0.83 AUC, HFmrEF 0.85 AUC and HFrEF 0.86 AUC, respectively). Conclusion Our findings present a novel, patient-level, AI-based risk prediction of all-cause mortality in heart failure with a robust accuracy across its phenotypes. This suggests the potential of AI based predictive models in a point-of-care approach to guide clinical risk stratification. Funding Acknowledgement Type of funding sources: Foundation. Main funding source(s): VZW Cardiovascular Research Center Aalst

Cardiac transplant-related vasculopathy remains a leading cause of morbidity and mortality in heart transplant (HTx) recipients. Recently, coronary angiography-derived vessel fractional flow reserve (vFFR) has emerged as a new diagnostic computational tool to functionally evaluate the severity of coronary artery disease. Although vFFR estimates have been shown to perform well against invasive FFR in atherosclerotic coronary artery disease, data on the use of vFFR in heart transplant recipients suffering from cardiac transplant-related arteriopathy are lacking. The aim of the presented study was to validate coronary angiography-derived vessel fractional flow reserve to calculate fractional flow reserve in HTx patients with and without cardiac transplant-related vasculopathy. A prospective, single center study of HTx patients referred for annual check-up, undergoing surveillance coronarography was conducted. Invasive FFR was measured using a motorized device at the speed of 1.0 mm/s in all three major coronary arteries. Angiography-derived pullback FFR was derived from the angiogram and compared with invasive FFR pullback curve. Overall, 18,059 FFR values were extracted from the FFR pullback curves from 23 HTx patients. The mean age was 59.3 ± 9.7 years, the mean time after transplantation was 5.24 years [IQR 1.20, 11.25]. A total of 39 vessels from 23 patients (24 LAD, 11 LCX, 4 RCA) were analyzed. Mean distal vFFR was 0.87 ± 0.14 whereas invasive distal FFR was 0.88 ± 0.17. An excellent correlation was found between invasive distal FFR and vFFR (r = 0.92; p < 0.001). The correlation of the pullback tracing was high, with a correlation coefficient between vFFR and invasive FFR pullback values of 0.72 (95% CI 0.71 to 0.73, p < 0.001). The mean difference between vFFR and invasive FFR pullback values was −0.01 with 0.06 of SD (limits of agreements −0.12 to 0.13). In HTx patients, coronary angiography-derived FFR correlates excellently with invasively measured wire-derived FFR. Therefore, angiography derived FFR could be used as a novel diagnostic tool to quantify the functional severity of graft vasculopathy.

Introduction: Heart failure (HF) is a heterogenous syndrome with complex pathophysiology. Biomarkers and clinical risk scores often fail to capture modifications in the treatment continuity and provide suboptimal patient-level precision in the prognostic stratification. Electronic patient records provide necessary granularity yielding opportunities to develop new artificial intelligence (AI) based strategies for comprehensive prognostic re-stratification. Hypothesis: We assessed the hypothesis that, utilizing longitudinal patient data in an AI approach, yields superior performance predicting all-cause mortality in a cohort of patients hospitalized with a de novo or worsened HF, compared to single observational time point predictions. Methods: In a cohort of 2449 HF patients hospitalized between 2011-2017, we utilized 151 451 patient exams from 422 parameters. Features included clinical phenotyping, medication, ECG, laboratory, echocardiography, catheterization data or percutaneous and surgical interventions gathered on a routine clinical basis reflecting standard of care as captured in individual electronic records. AI models were developed, and their performance on the validation set was compared to industry standard clinical scores. Results: AI models yielded performance ranging from 0.83 to 0.89 AUC on the outcome-balanced validation set in predicting all-cause mortality at 30-, 90-, 180-, 360- and 720-day time-limits. The primary endpoint, 1-year mortality prediction model, recorded 0.85 AUC on the validation set compared to 0.7 AUC (Seattle HF model) and 0.73 AUC (MAGGIC HF Score) respectively. Conclusions: Our findings present a novel, patient-level, AI-based risk prediction approach of all-cause mortality in heart failure utilizing all historical data available in electronic health records. This suggests the potential of AI based predictive models in a point-of-care approach to guide clinical risk stratification.