Machine learning-based prediction of three-year mortality in elderly inpatients with coronary artery disease combined with heart failure.

AimsInsulin‐like growth factor binding protein‐4 (IGFBP‐4) fragments have been shown to predict the risk of major adverse cardiovascular events, including segment‐elevation myocardial infarction, in patients with acute coronary syndrome. We evaluated the prognostic value of the carboxy‐terminal fragment of IGFBP‐4 (CT‐IGFBP‐4) for all‐cause mortality in emergency room patients with acute heart failure (AHF).Methods and resultsCT‐IGFBP‐4, N‐terminal pro brain natriuretic peptide (NT‐proBNP), and C‐reactive protein (CRP) were measured at admission from the lithium‐heparin plasma of 156 patients with AHF. All‐cause mortality was recorded for 1 year. Receiver operator characteristic (ROC) curves, Kaplan–Meier, and Cox proportional hazard ratio analyses were performed to evaluate the prognostic value of the various clinical variables, CT‐IGFBP‐4, NT‐proBNP, CRP, and their combinations. During 1 year of follow‐up, 52 (33.3%) patients died. CT‐IGFBP‐4 only weakly correlated with NT‐proBNP (Pearson correlation coefficient r = 0.16, P = 0.044) and did not correlate with CRP (r = 0.08, P = 0.35), emphasizing the different nature of these biomarkers. The receiver operator characteristic area under the curve (ROC AUC) of CT‐IGFBP‐4 for the prediction of all‐cause mortality (0.727) was significantly higher than that of NT‐proBNP (0.680, P = 0.045) and CRP (0.669, P = 0.016). The combination of CT‐IGFBP‐4, NT‐proBNP, and CRP predicted mortality significantly better (ROC AUC = 0.788) than any of the biomarkers alone (P < 0.01 for all). The addition of CT‐IGFBP‐4 to a clinical prediction model that included age, gender, systolic blood pressure, creatinine, and sodium levels, as well as the history of previous heart failure, coronary artery disease, and hypertension significantly improved the mortality risk prediction (ROC AUC 0.774 vs. 0.699, P = 0.025). Cox hazard analysis indicated that elevated CT‐IGFBP‐4 was independently associated with 1 year mortality (hazard ratio 3.26, P = 0.0008) after adjustment for age, gender, history of previous heart failure, coronary artery disease, hypertension, chronic kidney failure, history of diabetes, heart rate, haemoglobin, plasma sodium, NT‐proBNP, CRP, cystatin C, and elevated cardiac troponin I or T. Patients with increased levels of either two or three of the biomarkers CT‐IGFBP‐4, NT‐proBNP, and CRP had significantly higher mortality risk (adjusted hazard ratio 10.04, P < 0.0001) than patients with increased levels of one or none of the biomarkers.ConclusionsCT‐IGFBP‐4 was independently associated with all‐cause mortality in patients with AHF. Compared with single biomarkers, the combination of CT‐IGFBP‐4, NT‐proBNP, and CRP improved the prediction of all‐cause mortality in patients with AHF.

Is preventive medicine responsible for the increasing prevalence of heart failure?

Clinicians have become increasingly sophisticated in their application of cardiac biomarkers in the management of acute coronary syndromes (ACS). In the 1950s, clinical investigators first reported that proteins released from necrotic cardiac myocytes could be detected in the serum and could aid in the diagnosis of acute myocardial infarction.1 The ensuing 40 years witnessed progressive improvement in the cardiac tissue-specificity of biomarkers of myocardial necrosis and a corresponding enhancement in the clinical sensitivity and specificity of their use for establishing the diagnosis of acute myocardial infarction. Over the past decade, the emergence of convincing evidence for the value of cardiac troponin in guiding therapy has dramatically accelerated the integration of cardiac biomarkers into clinical decision-making for patients with ACS.2 Concurrently, advances in our understanding of the pathogenesis and consequences of acute coronary atherothrombosis have stimulated the development of new biomarkers and created the opportunity for an expanded role of multiple biomarkers, some old and others new, in the classification and individualization of treatment for ACS.3,4 The report by James et al5 in the present issue of Circulation adds substantially to the accumulating evidence that a multimarker strategy, employing a pathobiologically diverse set of biomarkers,3 is likely to add importantly to cardiac-specific troponin alone in the risk assessment of patients with ACS. See p 275 ACS is a complex syndrome with multiple causes, analogous to anemia or hypertension.6 As such, treatment is likely to be most effective when directed at the underlying cause of the disease. Five principal causes of ACS have been described; these include (1) plaque rupture with acute thrombosis, (2) progressive mechanical obstruction, (3) inflammation, (4) secondary unstable angina, and (5) dynamic obstruction (coronary vasoconstriction).7 It is rare that any of these contributors exists in isolation. However, patients with ACS may …

Brain (B-type) natriuretic peptide (BNP) is a 32 amino acid peptide that is synthesized and released predominantly from ventricular myocardium in response to myocyte stretch. Like atrial natriuretic peptide (ANP), BNP seems to have almost exclusively beneficial physiological properties, including balanced vasodilation, natriuresis, and inhibition of both the sympathetic nervous system and the renin-angiotensin-aldosterone axis. Attempts to exploit these properties for therapeutic benefit has led to the development of recombinant human BNP (nesiritide) for the acute treatment of decompensated heart failure, and also of novel compounds that inhibit neutral endopeptidase, an enzyme that is partially responsible for BNP degradation. See p 2913 In patients with heart failure, the cardiac neurohormonal system is activated, and circulating plasma levels of ANP, BNP, and the N-terminal fragments of their prohormones (N-proANP and N-proBNP) are elevated. Compared with ANP and N-proANP, BNP and N-proBNP undergo a greater proportional rise in disease states (ie, higher “signal-to-noise” ratio), and thus have emerged as the preferred biomarkers for clinical development. With commercially available assays now available, measurement of BNP or N-proBNP can be integrated readily into the care of patients with suspected heart failure. Although data are limited, BNP and N-proBNP seem to provide qualitatively similar information, and for purposes of this editorial, will be referred to interchangeably. Incorporation of BNP measurement into the clinical evaluation facilitates the diagnosis of heart failure due to either left ventricular (LV) systolic or diastolic dysfunction; a normal BNP level virtually rules out the diagnosis of decompensated heart failure, whereas a markedly elevated BNP has a high positive predictive value for heart failure.1 Although BNP levels are correlated with age, sex, intracardiac filling pressures, LV mass and ejection fraction (LVEF), renal function, and symptoms, BNP provides prognostic information in patients with heart failure that is independent of these variables.2 …

Plasma brain-type natriuretic peptide (BNP) and amino-terminal proBNP (NTproBNP) provide prognostic information on cardiovascular morbidity and mortality beyond that provided by standard risk factors. Clinical applications of B-type peptides under ongoing research include their use in diagnosing acute heart failure (HF), in risk stratification in both acute and established HF, in acute coronary syndromes (ACS), in asymptomatic populations at cardiovascular risk (older adults and people with hypertension), and as part of a screening strategy for detection of left ventricular impairment and prediction of cardiovascular risk in the general population.1,2 In this issue of Circulation , Campbell and colleagues3 assess the ability of NTproBNP to predict myocardial infarction (MI) in subjects who have experienced a cerebrovascular event. NTproBNP (reflecting cardiac distension) is compared with C-reactive protein (a systemic marker of inflammation) and renin (a marker of sodium status regulated by renal perfusion and delivery of sodium to the renal glomerulus). See p 110 The nested case-control study is from the 6105 participants in the Perindopril Protection Against Recurrent Stroke Study (PROGRESS), a placebo-controlled study of converting enzyme inhibitor–based therapy in patients with previous cerebrovascular events.4 Within PROGRESS, 206 subjects incurred an MI during 3.9 years of follow-up. The investigators matched those incurring an MI with control PROGRESS patients avoiding MI from time of randomization to time of case ascertainment. Cases and controls were matched for age, gender, treatment allocation, region, and cerebrovascular qualifying event. The form of matching meant that individual patients may have been controls initially and subsequently became cases on incurring an MI during further follow-up. Matching in this fashion may confuse the interpretation of the nonconditioned analysis of baseline variables. Comparing the 206 cases with 412 controls at randomization, the investigators report that in addition to higher systolic blood pressure, more frequent known coronary disease, …

Patients with end-stage renal disease (ESRD) have an increased mortality from cardiovascular disease (CVD). N-terminal pro-brain natriuretic peptide (NT-pro-BNP) is an independent predictor of mortality in patients with ischemic heart disease and congestive heart failure. Previous data have shown markedly elevated levels of NT-pro-BNP in patients with ESRD, while the prognostic value of elevated levels of NT-pro-BNP in patients with ESRD is largely unknown. The aim of the present study was to examine if the level of NT-pro-BNP predicts mortality in patients with ERSD and CVD. We prospectively followed 206 patients with ESRD and documented CVD. Levels of NT-pro-BNP were measured at baseline, and patients were followed for 2 years or until they reached the predefined endpoint of all-cause mortality. During follow-up, the total mortality was 44% (90/206). Patients who died were followed for a median of 314 days (interquartile range 179 - 530). Using Cox regression analysis, age, female sex, systolic blood pressure, dialysis efficiency and plasma levels of NT-pro-BNP were independent prognostic risk factors of mortality. In receiver operating characteristic curve analysis a cut off value for NT-pro-BNP was determined. Patients with values of NT-pro-BNP above 12.200 pg/ml had a 3 times higher risk of death than patients below the cut-off value (HR 3.05 95% CI 1.96 - 4.77, p < 0.0001). In spite of generally elevated levels of NT-pro-BNP, NT-pro-BNP is still an independent predictor of mortality and might add prognostic information in patients with ESRD and documented CVD.

O verweight and obesity have become increasingly com- mon; worldwide, at least 1.1 billion adults are overweight and 312 million are obese, when overweight and obesity are defined conventionally as having a body mass index (BMI) of Ͼ25 kg/m 2 and Ͼ30 kg/m 2 , respectively. 1,2In the general population, overweight and obesity are associated with increased risk of developing cardiovascular disease, 3,4 and thus it is not surprising that in cohorts of patients with prevalent ischemic heart disease or acute coronary events, well over 50% are overweight or obese. 5,6Despite the association between obesity and cardiovascular risk in the general population, a multitude of studies have described an inverse correlation between BMI and mortality in patients with coronary artery disease (CAD), including post-coronary revascularization patients and those with acute myocardial infarction (MI); the association between elevated BMI and improved survival has been termed the obesity paradox. 7,8 Article p 482In this issue of Circulation, Zeller et al 9 further investigate the obesity paradox in a cohort of 2229 consecutive patients presenting with acute MI in the Côte d'Or region of France.In assessing the impact of obesity on mortality after MI, the group uses both BMI, a traditional index of obesity, as well as waist circumference, an alternate anthropometric index that is more specific for abdominal obesity.Approximately one-half of the subjects in the study were overweight (BMI 25 to 29.9 kg/m 2 ), one-quarter were obese (BMI Ͼ30 kg/m 2 ) and onehalf had increased waist circumference, which was defined as Ͼ102 cm in men and Ͼ88 cm in women.Left ventricular ejection fraction, type of MI, and acute treatment strategy did not generally differ by BMI or waist circumference values.Of note, BMI was inversely correlated with age and plasma N-terminal pro B-type natriuretic peptide, whereas waist circumference was positively correlated with age and did not correlate with N-terminal pro B-type natriuretic peptide.Consistent with prior studies, survival analysis showed that the risk of death decreased with increasing BMI tertile.In a waist-matched analysis of 832 subjects, BMI was a signifi-

N-Terminal Pro–B-Type Natriuretic Peptide-Guided Treatment for Chronic Heart Failure: Results From the BATTLESCARRED (NT-proBNP–Assisted Treatment To Lessen Serial Cardiac Readmissions and Death) Trial

Cardiac myosin-binding protein C as a candidate biomarker in heart failure: rational but not revolutionary.

An investigation of machine learning methods applied to genomic prediction in yellow-feathered broilers

ORAL PRESENTATION

N-Terminal Pro-B-Type Natriuretic Peptide as an Indicator of Disease Severity in a Heterogeneous Group of Patients With Chronic Precapillary Pulmonary Hypertension

ObjectiveTo analyze the relationship between N-terminal pro-brain natriuretic peptide (NT-proBNP) and renal function, and compare the ability and cut-off thresholds of NT-proBNP to detect chronic heart failure (CHF) and predict mortality in elderly Chinese coronary artery disease (CAD) patients with and without chronic kidney disease (CKD).MethodsThe study included 999 CAD patients older than 60 years. The endpoint was all-cause mortality over a mean follow-up period of 417 days.ResultsThe median age was 86 years (range: 60–104 years), and the median NT-proBNP level was 409.8 pg/mL. CKD was present in 358 patients. Three hundred and six patients were positive for CHF. One hundred and ten CKD patients and 105 non-CKD patients died. Not only CKD, but also estimated glomerular filtration rate independently affected NT-proBNP. NT-proBNP detected CHF with a cut-off value of 298.4 pg/mL in non-CKD patients and a cut-off value of 435.7 pg/mL in CKD patients. NT-proBNP predicted death with a cut-off value of 369.5 pg/mL in non-CKD patients and a cut-off value of 2584.1 pg/mL in CKD patients. The NT-proBNP level was significantly related to the prevalence of CHF and all-cause mortality in CAD patients with and without CKD; this effect persisted after adjustment. The crude and multiple adjusted hazard ratios of NT-proBNP to detect CHF and predict mortality were significantly higher in patients with CKD compared with the remainder of the population. The addition of NT-proBNP to the three-variable and six-variable models generated a significant increase in the C-statistic.ConclusionAmongst elderly Chinese CAD patients, there was an independently inverse association between NT-proBNP and renal function. With the higher cutoff points, NT-proBNP better detected CHF and better predicted mortality in CKD patients than in non-CKD patients.

Heart failure (HF) is a growing public health problem in the United States. Nearly 5 million Americans suffer from HF, and an estimated 550 000 new cases of HF are diagnosed each year.1 HF is the No. 1 discharge diagnosis in patients ≥65 years of age and results in a substantial burden on healthcare expenditures. It is estimated that in 2001, more than $24 billion was spent as direct cost for the care of patients with HF.1 Furthermore, HF is associated with a significant increase in morbidity and mortality. See p 1764 Although considerable progress has been made in our approach to the pharmacological management of patients with HF, most patients remain at increased risk of cardiac death. To further improve outcomes in patients with HF, newer therapeutic modalities, including devices such as biventricular pacemaker, automatic internal cardioverter-defibrillators (AICDs), and left ventricular assist devices, have been increasingly utilized. Several recent randomized controlled trials have shown that such devices can indeed further improve the outcome in patients with HF.2–4 However, these devices are expensive, and their widespread or injudicious application in unselected patients with HF is likely to have a substantial impact on healthcare expenditures. On the other hand, appropriate use of device therapy in properly selected patients (who are at high risk of mortality) is essential to improve clinical outcome. Thus, there is a need to develop a strategy to accurately identify those patients with HF who are at increased risk of mortality. The paper by Vrtovec and associates5 in the present issue of Circulation provides such a strategy by showing that routinely available diagnostic tests, such as measurement of QT interval on 12-lead ECG and measurement of B-type natriuretic peptide (BNP), can indeed identify the HF patients who are at increased risk of overall mortality, …

Prediction of mortality in heart failure by machine learning. Comparison with statistical modeling.