Redefining β-blocker response in heart failure patients with sinus rhythm and atrial fibrillation: a machine learning cluster analysis

Abstract

Background: Mortality remains unacceptably high in patients with heart failure and reduced left ventricular ejection fraction (LVEF) despite advances in therapeutics. We hypothesised that a novel artificial intelligence approach could better assess multiple and higher-dimension interactions of comorbidities, and define clusters of β-blocker efficacy in patients with sinus rhythm and atrial fibrillation.

Methods: Neural network-based variational autoencoders and hierarchical clustering were applied to pooled individual patient data from nine double-blind, randomised, placebo-controlled trials of β blockers. All-cause mortality during median 1·3 years of follow-up was assessed by intention to treat, stratified by electrocardiographic heart rhythm. The number of clusters and dimensions was determined objectively, with results validated using a leave-one-trial-out approach. This study was prospectively registered with ClinicalTrials.gov (NCT00832442) and the PROSPERO database of systematic reviews (CRD42014010012).

Findings: 15 659 patients with heart failure and LVEF of less than 50% were included, with median age 65 years (IQR 56-72) and LVEF 27% (IQR 21-33). 3708 (24%) patients were women. In sinus rhythm (n=12 822), most clusters demonstrated a consistent overall mortality benefit from β blockers, with odds ratios (ORs) ranging from 0·54 to 0·74. One cluster in sinus rhythm of older patients with less severe symptoms showed no significant efficacy (OR 0·86, 95% CI 0·67-1·10; p=0·22). In atrial fibrillation (n=2837), four of five clusters were consistent with the overall neutral effect of β blockers versus placebo (OR 0·92, 0·77-1·10; p=0·37). One cluster of younger atrial fibrillation patients at lower mortality risk but similar LVEF to average had a statistically significant reduction in mortality with β blockers (OR 0·57, 0·35-0·93; p=0·023). The robustness and consistency of clustering was confirmed for all models (p<0·0001 vs random), and cluster membership was externally validated across the nine independent trials.

Interpretation: An artificial intelligence-based clustering approach was able to distinguish prognostic response from β blockers in patients with heart failure and reduced LVEF. This included patients in sinus rhythm with suboptimal efficacy, as well as a cluster of patients with atrial fibrillation where β blockers did reduce mortality.

Funding: Medical Research Council, UK, and EU/EFPIA Innovative Medicines Initiative BigData@Heart.

Conflict of interest statement

Declaration of interests AK reports a grant from the UK Medical Research Council (MRC; MR/S003991/1), outside of the study. KVB reports a grant from the University of Birmingham/British Heart Foundation (BHF) Accelerator Award (BHF AA/18/2/34218). SKG is funded through the EU/EFPIA Innovative Medicines Initiative (BigData@Heart 116074), outside the submitted work. OT is funded from the EU/EFPIA Innovative Medicines Initiative (BigData@Heart 116074) and Amomed Pharma, outside the submitted work. JD reports grants from BHF Accelerator Award (BHF AA/18/2/34218), outside the submitted work; in addition, they have a patent pending (WO2020070519; method for detecting adverse cardiac events). LS reports grants from Nanocommons and Health Data Research (HDR) UK, outside the submitted work. AJSC reports personal fees from Menarini, AstraZeneca, Bayer, Boehringer Ingelheim, Novartis, Nutricia, Servier, Vifor, Abbott, Actimed, Arena, Cardiac Dimensions, Corvia, CVRx, Enopace, ESN Cleer, Faraday, Gore, Impulse Dynamics, and Respicardia, all outside the submitted work. GVG reports support from the National Institute for Health Research (NIHR) Birmingham Experimental Cancer Medicine Centre, NIHR Birmingham Surgical Reconstruction and Microbiology Research Centre, Nanocommons H2020-EU (731032), and the MRC Heath Data Research UK (HDRUK/CFC/01). DK reports grants from the NIHR (NIHR CDF-2015-08-074 RATE-AF; NIHR HTA-130280 DaRe2THINK; NIHR EME-132974 D2T-NV); BHF (PG/17/55/33087 and AA/18/2/34218); EU/EFPIA Innovative Medicines Initiative (BigData@Heart 116074); European Society of Cardiology supported by educational grants from Boehringer Ingelheim, Bristol Myers Squibb-Pfizer Alliance, Bayer, Daiichi Sankyo, and Boston Scientific; NIHR/University of Oxford Biomedical Research Centre and University of Birmingham/BHF Accelerator Award (STEEER-AF NCT04396418); and Amomed Pharma and IRCCS San Raffaele/Menarini (Beta-blockers in Heart Failure Collaborative Group NCT0083244). DK also declares personal fees from Bayer (advisory board), AtriCure (speaker fees), Amomed (advisory board), Protherics Medicines Development (advisory board), and Myokardia (advisory board). SP, FA, ADB, SC, JD, ARM, VRC, JAW, E-JB, XW and MDF declare no competing interests.

Copyright © 2021 The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY 4.0 license. Published by Elsevier Ltd.. All rights reserved.

Figures

Figure 1

Study flowchart ECG=electrocardiogram. LVEF=left ventricular ejection fraction. RCTs=randomised controlled trials. *Age, gender, body-mass index (missing n=148), heart rate (missing n=5), systolic blood pressure (missing n=7), LVEF, previous myocardial infarction (missing n=43), New York Heart Association symptom class (missing n=83), creatinine (missing n=101), baseline drug therapy (angiotensin converting enzyme inhibitor or angiotensin receptor blocker, diuretics, anticoagulants [missing n=1] and digoxin), and other than sinus rhythm or atrial fibrillation or flutter on baseline ECG (n=506). 32 patients had more than one variable missing. † MDC, US-HF, ANZ, CIBIS-II, MERIT-HF, COPERNICUS, CAPRICORN, BEST, and SENIORS (appendix p 16).

Figure 2

Clustering for all-cause mortality and β-blocker efficacy in SR Green circles represent the average mortality risk, with size relative to the number of patients in that cluster. ORs (95% CI) are for the efficacy of β blockers versus placebo for all-cause mortality; odds below the dotted line indicate a benefit from β blockers. Radar plots summarise scaled variables for each cluster, with the average for the whole cohort of sinus rhythm patients noted in orange. Values closer to the outer ring are higher than the cohort average for each of the key variables. Other variables not displayed in the radar plots include: systolic blood pressure, New York Heart Association symptom class, creatinine, and baseline drug therapy (appendix p 9). OR=odds ratio. BMI=body-mass index. HR=heart rate. LVEF=left ventricular ejection fraction. MI=myocardial infarction. SR=sinus rhythm.

Figure 3

Clustering for all-cause mortality and β-blocker efficacy in AF Blue circles represent the average mortality risk, with size relative to the number of patients in that cluster. ORs (95% CI) are for the efficacy of β blockers versus placebo for all-cause mortality; odds below the dotted line indicate a benefit from β blockers. Radar plots summarise scaled variables for each cluster, with the average for the whole cohort of atrial fibrillation noted in orange. Values closer to the outer ring are higher than the cohort average for each of the key variables. Other variables not displayed in the radar plots include: systolic blood pressure, New York Heart Association symptom class, creatinine, and baseline drug therapy (appendix p 11). OR=odds ratio. AF=atrial fibrillation. BMI=body-mass index. HR=heart rate. LVEF=left ventricular ejection fraction. MI=myocardial infarction.