Diagnostic imaging modalities for acute heart failure: a systematic review with network meta-analysis
A S Olesen, K Miger, I A Taraldsen, J Lukoschewitz, E Melgaard, S Tonning, L Westergaard, S Haastrup, M El-Sheikh, O W Nielsen, J J Thune, P S Jhund, A D Henderson, J C Jakobsen, J GrandAbstract
Background
Rapid identification of pulmonary congestion in patients with acute dyspnoea is central to early evaluation and management of acute heart failure. In emergency settings, diagnostic decisions often rely on available imaging modalities, but their comparative performance for detecting acute heart failure remains uncertain.
Purpose
To compare the diagnostic accuracy of chest radiography, lung ultrasound, computed tomography (CT), and remote dielectric sensing (ReDS) for detecting pulmonary congestion indicative of acute heart failure in adults with dyspnoea, and to evaluate methodological quality.
Methods
This systematic review and meta-analysis followed PRISMA and Cochrane guidance. It was registered in PROSPERO, and a prior protocol was published. Medline, Embase, Latin American and Caribbean Health Sciences Literature, and Web of Science were searched on 9 April 2025.
We included prospective observational studies of adults with acute dyspnoea comparing at least one modality against an investigator-defined reference standard for pulmonary congestion or acute heart failure. Two reviewers independently screened studies, extracted data, and assessed risk of bias. Sensitivities and specificities were pooled using bivariate random-effects models. A Bayesian network meta-analysis with weakly informative priors was performed to compare modalities.
Results
Of 156,101 studies screened, 95 studies comprising 15,132 patients were included. Among these, 78 studies evaluated lung ultrasound, 30 chest radiography, six ReDS, and two CT. Bivariate pooled sensitivity and specificity were 0.87 (95%CI 0.82-0.90) and 0.89 (0.85-0.91) for lung ultrasound; 0.65 (0.58-0.72) and 0.91 (0.87-0.94) for chest radiography; and 0.73 (0.53-0.87) and 0.73 (0.58-0.84) for ReDS (Picture 1). CT was not included due to limited data.
In Bayesian meta-analysis (Picture 2), CT and lung ultrasound showed an >95% posterior probability of higher sensitivity than chest radiography (CT: 0.76 vs 0.62; 95%CrI for difference 0.01-0.24; lung ultrasound: 0.86 vs 0.62; 95%CrI 0.19-0.28). CT demonstrated higher specificity than chest radiography (0.94 vs 0.85; 95%CrI 0.03-0.13) and ReDS (0.94 vs 0.73; 95%CrI 0.06-0.40), while lung ultrasound also had higher specificity than chest radiography (0.89 vs 0.85; 95%CrI 0.02-0.06) and ReDS (0.89 vs 0.73; 95%CrI 0.03-0.36).
Only four studies were rated as overall low risk of bias, reflecting heterogeneity in reference standards and patient selection.
Conclusion
Lung ultrasound demonstrates the most favourable balance of diagnostic performance and bedside availability for detecting acute heart failure in patients with dyspnoea. CT is useful when lung ultrasound is inconclusive, or when other pathology is also suspected. Chest radiography may support rule-in but cannot exclude acute heart failure when normal. ReDS showed moderate diagnostic accuracy.Summary ROC curves from bivariate modelsFor image description, please refer to the figure legend and surrounding text.Posterior sensitivitiy and specificityFor image description, please refer to the figure legend and surrounding text.