Tang, W H Wilson

Abstract Deep learning (DL) is a branch of machine learning (ML) showing increasing promise in medicine, to assist in data classification, novel disease phenotyping and complex decision making. Deep learning is a form of ML typically implemented via multi-layered neural networks. Deep learning has accelerated by recent advances in computer hardware and algorithms and is increasingly applied in e-commerce, finance, and voice and image recognition to learn and classify complex datasets. The current medical literature shows both strengths and limitations of DL. Strengths of DL include its ability to automate medical image interpretation, enhance clinical decision-making, identify novel phenotypes, and select better treatment pathways in complex diseases. Deep learning may be well-suited to cardiovascular medicine in which haemodynamic and electrophysiological indices are increasingly captured on a continuous basis by wearable devices as well as image segmentation in cardiac imaging. However, DL also has significant weaknesses including difficulties in interpreting its models (the ‘black-box’ criticism), its need for extensive adjudicated (‘labelled’) data in training, lack of standardization in design, lack of data-efficiency in training, limited applicability to clinical trials, and other factors. Thus, the optimal clinical application of DL requires careful formulation of solvable problems, selection of most appropriate DL algorithms and data, and balanced interpretation of results. This review synthesizes the current state of DL for cardiovascular clinicians and investigators, and provides technical context to appreciate the promise, pitfalls, near-term challenges, and opportunities for this exciting new area.

Purpose: Psychosocial assessment of patients comprises an important element in the selection process of appropriate candidates for left ventricular assist device (LVAD) implantation. We sought to determine the association of the well-validated psychosocial assessment of candidates for transplantation (PACT) scale to clinical outcomes post-LVAD implantation. Materials and Methods: The PACT scale was used retrospectively to reconstruct psychosocial profiles of all patients who underwent a continuous-flow LVAD implantation for all indications at our institution between March 2008 and August 2012 (N = 230). Psychosocial elements including social support, psychological health, lifestyle factors, comprehension of the operation, and follow-up were evaluated. The primary outcome was overall survival, and the secondary outcomes were hospital readmission, pump thrombosis, hemolysis, gastrointestinal (GI) bleeding, and LVAD driveline infections. Results: The mean age of patients was 55.3 years, with 83% being male; 58% (N = 135) were bridge to transplant and 42% (N = 95) were destination therapy. Up to 1-year post-LVAD implant, there were no statistical differences among the 5 PACT candidate groups in terms of survival ( P = .79), hospital readmissions ( P = .55), suspected or confirmed pump thrombosis ( P = .31), hemolysis ( P = .43), GI bleeding ( P = .71), or driveline infections ( P = .06). Conclusions: In this single-center retrospective review, post hoc reconstruction of psychosocial profiles using the PACT scale and independent assessment of postimplant outcomes, including survival and adverse events, did not show any association. However, given the small number of patients in the low score PACT groups as well as limited duration of follow-up, further studies are required to elucidate the association.

Background: Worsening renal function (WRF) in the setting of aggressive diuresis for acute heart failure treatment may reflect renal tubular injury or simply indicate a hemodynamic or functional change in glomerular filtration. Well-validated tubular injury biomarkers, N -acetyl-β- d -glucosaminidase, neutrophil gelatinase-associated lipocalin, and kidney injury molecule 1, are now available that can quantify the degree of renal tubular injury. The ROSE-AHF trial (Renal Optimization Strategies Evaluation–Acute Heart Failure) provides an experimental platform for the study of mechanisms of WRF during aggressive diuresis for acute heart failure because the ROSE-AHF protocol dictated high-dose loop diuretic therapy in all patients. We sought to determine whether tubular injury biomarkers are associated with WRF in the setting of aggressive diuresis and its association with prognosis. Methods: Patients in the multicenter ROSE-AHF trial with baseline and 72-hour urine tubular injury biomarkers were analyzed (n=283). WRF was defined as a ≥20% decrease in glomerular filtration rate estimated with cystatin C. Results: Consistent with protocol-driven aggressive dosing of loop diuretics, participants received a median 560 mg IV furosemide equivalents (interquartile range, 300–815 mg), which induced a urine output of 8425 mL (interquartile range, 6341–10 528 mL) over the 72-hour intervention period. Levels of N -acetyl-β- d -glucosaminidase and kidney injury molecule 1 did not change with aggressive diuresis (both P >0.59), whereas levels of neutrophil gelatinase-associated lipocalin decreased slightly (−8.7 ng/mg; interquartile range, −169 to 35 ng/mg; P <0.001). WRF occurred in 21.2% of the population and was not associated with an increase in any marker of renal tubular injury: neutrophil gelatinase-associated lipocalin ( P =0.21), N -acetyl-β- d -glucosaminidase ( P =0.46), or kidney injury molecule 1 ( P =0.22). Increases in neutrophil gelatinase-associated lipocalin, N -acetyl-β- d -glucosaminidase, and kidney injury molecule 1 were paradoxically associated with improved survival (adjusted hazard ratio, 0.80 per 10 percentile increase; 95% confidence interval, 0.69–0.91; P =0.001). Conclusions: Kidney tubular injury does not appear to have an association with WRF in the context of aggressive diuresis of patients with acute heart failure. These findings reinforce the notion that the small to moderate deteriorations in renal function commonly encountered with aggressive diuresis are dissimilar from traditional causes of acute kidney injury.

Background— The ROSE AHF trial (Renal Optimization Strategies Evaluation in Acute Heart Failure) found that when compared with placebo, neither low-dose dopamine (2 µg/kg per minute) nor low-dose nesiritide (0.005 μg/kg per minute without bolus) enhanced decongestion or preserved renal function in AHF patients with renal dysfunction. However, there may be differential responses to vasoactive agents in AHF patients with reduced versus preserved ejection fraction (EF). This post hoc analysis examined potential interaction between treatment effect and EF (EF ≤40% versus >40%) on the ROSE AHF end points. Methods and Results— ROSE AHF enrolled AHF patients (n=360; any EF) with renal dysfunction. The coprimary end points were cumulative urine volume and the change in serum cystatin-C in 72 hours. The effect of dopamine (interaction P =0.001) and nesiritide (interaction P =0.039) on urine volume varied by EF group. In heart failure with reduced EF, urine volume was higher with active treatment versus placebo, whereas in heart failure with preserved EF, urine volume was lower with active treatment. The effect of dopamine and nesiritide on weight change, sodium excretion, and incidence of AHF treatment failure also varied by EF group (interaction P <0.05 for all). There was no interaction between vasoactive treatment’s effect and EF on change in cystatin-C. Compared with placebo, dopamine was associated with improved clinical outcomes in heart failure with reduced EF and worse clinical outcomes in heart failure with preserved EF. With nesiritide, there were no differences in clinical outcomes when compared with placebo in both heart failure with reduced EF and heart failure with preserved EF. Conclusions— In this post hoc analysis of ROSE AHF, the response to vasoactive therapies differed in patients with heart failure with reduced EF and heart failure with preserved EF. Investigations of AHF therapies should assess the potential for differential responses in AHF with preserved versus reduced EF. Clinical Trial Registration— URL: http://www.clinicaltrials.gov . Unique identifier: NCT01132846.

Background— Oxidative stress may contribute to heart failure (HF) progression. Inhibiting xanthine oxidase in hyperuricemic HF patients may improve outcomes. Methods and Results— We randomly assigned 253 patients with symptomatic HF, left ventricular ejection fraction ≤40%, and serum uric acid levels ≥9.5 mg/dL to receive allopurinol (target dose, 600 mg daily) or placebo in a double-blind, multicenter trial. The primary composite end point at 24 weeks was based on survival, worsening HF, and patient global assessment. Secondary end points included change in quality of life, submaximal exercise capacity, and left ventricular ejection fraction. Uric acid levels were significantly reduced with allopurinol in comparison with placebo (treatment difference, –4.2 [–4.9, –3.5] mg/dL and –3.5 [–4.2, –2.7] mg/dL at 12 and 24 weeks, respectively, both P <0.0001). At 24 weeks, there was no significant difference in clinical status between the allopurinol- and placebo-treated patients (worsened 45% versus 46%, unchanged 42% versus 34%, improved 13% versus 19%, respectively; P =0.68). At 12 and 24 weeks, there was no significant difference in change in Kansas City Cardiomyopathy Questionnaire scores or 6-minute walk distances between the 2 groups. At 24 weeks, left ventricular ejection fraction did not change in either group or between groups. Rash occurred more frequently with allopurinol (10% versus 2%, P =0.01), but there was no difference in serious adverse event rates between the groups (20% versus 15%, P =0.36). Conclusions— In high-risk HF patients with reduced ejection fraction and elevated uric acid levels, xanthine oxidase inhibition with allopurinol failed to improve clinical status, exercise capacity, quality of life, or left ventricular ejection fraction at 24 weeks. Clinical Trial Registration— URL: http://www.clinicaltrials.gov . Unique identifier: NCT00987415.