Artificial intelligence in cardiovascular care: a systematic review and meta-analysis of randomised controlled trials
File(s)
Published version
Author(s)
Type
Journal Article
Abstract
Background: Artificial intelligence (AI) holds potential to transform cardiovascular care, but evidence on its effectiveness in clinical practice remains inconsistent. We aimed to synthesise evidence from randomised controlled trials (RCTs) on the effectiveness of AI-enabled cardiovascular care, summarise the trial design and characteristics of AI systems, and evaluate methodological quality and reporting transparency.
Methods: In this systematic review and meta-analysis, we searched Embase, MEDLINE, Scopus, Cochrane Central Register of Controlled Trials, and ClinicalTrials.gov for RCTs that evaluated effectiveness of AI interventions in cardiovascular care, published in English from database inception to July 07, 2025. The search was updated on April 28, 2026. We followed Cochrane guidance for study selection and data extraction. Risk of bias was assessed using Cochrane’s Risk of Bias tools and reporting transparency using CONSORT-AI checklist. We calculated summary effects using inverse-variance-weighted random-effects meta-analyses and assessed the certainty of evidence using GRADE. Between-study heterogeneity was quantified using χ² (Cochran’s Q) test and I² statistic. Publication bias was not assessed due to small number of studies. This study was registered with PROSPERO (CRD420251090250).
Findings: Of 12,217 records identified, 31 RCTs from 13 regions (n=1,685,717 patients) were included in the systematic review, and 11 of these (n=1,614,689) in the meta-analysis. Most RCTs were published after 2021 (90%), multicentre (58%), and had short follow-up duration (<12 months; 52%). Risk of bias was low in seven trials (23%), and overall reporting transparency was moderate. Twenty-two trials (71%) reported significant benefit of AI interventions on primary endpoints, mostly intermediate process measures, while nine trials (29%) found no significant effect. Compared with routine care, image-based AI-clinical decision support system had significant effect on major adverse cardiovascular events (risk ratio [RR] 0.74 [95% CI 0.58–0.96]; I²=0%). AI-based mobile health interventions showed reduction in systolic blood pressure (mean difference -3.18 mm Hg [95% CI -5.50 to -0.86]; I²=0%), but not in diastolic blood pressure and body mass index. AI-enhanced electrocardiograms did not have significant effect on heart failure detection (RR 1.22 [0.70–2.14]; I²=90%). AI had unclear effects on atrial fibrillation diagnosis and echocardiogram utilisation due to very low certainty evidence. Certainty of evidence ranged from low to very low across outcomes due to risk of bias, indirectness, and imprecision.
Interpretation: AI demonstrated potential benefits in intermediate process measures in cardiovascular care but evidence of its effect on hard clinical outcomes remains limited. Current evidence is constrained by high risk of bias, trial heterogeneity, inadequate reporting of AI-specific components, limited generalisability, and a scarcity of trials with clinically meaningful endpoints. More rigorous, representative, and transparently reported RCTs with longer follow-up and long-term clinical outcomes are needed.
Funding: None received.
Methods: In this systematic review and meta-analysis, we searched Embase, MEDLINE, Scopus, Cochrane Central Register of Controlled Trials, and ClinicalTrials.gov for RCTs that evaluated effectiveness of AI interventions in cardiovascular care, published in English from database inception to July 07, 2025. The search was updated on April 28, 2026. We followed Cochrane guidance for study selection and data extraction. Risk of bias was assessed using Cochrane’s Risk of Bias tools and reporting transparency using CONSORT-AI checklist. We calculated summary effects using inverse-variance-weighted random-effects meta-analyses and assessed the certainty of evidence using GRADE. Between-study heterogeneity was quantified using χ² (Cochran’s Q) test and I² statistic. Publication bias was not assessed due to small number of studies. This study was registered with PROSPERO (CRD420251090250).
Findings: Of 12,217 records identified, 31 RCTs from 13 regions (n=1,685,717 patients) were included in the systematic review, and 11 of these (n=1,614,689) in the meta-analysis. Most RCTs were published after 2021 (90%), multicentre (58%), and had short follow-up duration (<12 months; 52%). Risk of bias was low in seven trials (23%), and overall reporting transparency was moderate. Twenty-two trials (71%) reported significant benefit of AI interventions on primary endpoints, mostly intermediate process measures, while nine trials (29%) found no significant effect. Compared with routine care, image-based AI-clinical decision support system had significant effect on major adverse cardiovascular events (risk ratio [RR] 0.74 [95% CI 0.58–0.96]; I²=0%). AI-based mobile health interventions showed reduction in systolic blood pressure (mean difference -3.18 mm Hg [95% CI -5.50 to -0.86]; I²=0%), but not in diastolic blood pressure and body mass index. AI-enhanced electrocardiograms did not have significant effect on heart failure detection (RR 1.22 [0.70–2.14]; I²=90%). AI had unclear effects on atrial fibrillation diagnosis and echocardiogram utilisation due to very low certainty evidence. Certainty of evidence ranged from low to very low across outcomes due to risk of bias, indirectness, and imprecision.
Interpretation: AI demonstrated potential benefits in intermediate process measures in cardiovascular care but evidence of its effect on hard clinical outcomes remains limited. Current evidence is constrained by high risk of bias, trial heterogeneity, inadequate reporting of AI-specific components, limited generalisability, and a scarcity of trials with clinically meaningful endpoints. More rigorous, representative, and transparently reported RCTs with longer follow-up and long-term clinical outcomes are needed.
Funding: None received.
Date Issued
2026-08-01
Date Acceptance
2026-07-16
Citation
EClinicalMedicine, 2026, 98
ISSN
2589-5370
Publisher
Elsevier
Journal / Book Title
EClinicalMedicine
Volume
98
Copyright Statement
© 2026 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
License URL
Subjects
Artificial intelligence
Machine learning
Cardiovascular disease
Cardiology
Clinical trial
Evidence synthesis
Publication Status
Published
Article Number
104123
Date Publish Online
2026-08-06