Steffen Tiedt1,2, Stefan Brandmaier3,4, Hanna Kollmeier1, Marco Duering1,2, Anna Artati5, Jerzy Adamski5,6,7,8, Matthias Klein9, Thomas Liebig10, Lesca M Holdt11, Daniel Teupser11, Rui Wang-Sattler3,4,7, Edzard Schwedhelm12,13, Christian Gieger3,4,7, Martin Dichgans1,2. 1. Institute for Stroke and Dementia Research, University Hospital, LMU Munich, Munich, Germany. 2. Munich Cluster for Systems Neurology (SyNergy), Munich, Germany. 3. Research Unit of Molecular Epidemiology, Helmholtz Zentrum München, Neuherberg, Germany. 4. Institute of Epidemiology, Helmholtz Zentrum München, Neuherberg, Germany. 5. Research Unit Molecular Endocrinology and Metabolism, Genome Analysis Center, Helmholtz Zentrum München, Neuherberg, Germany. 6. Institute of Experimental Genetics, Technical University of Munich, Freising, Germany. 7. German Center for Diabetes Research (DZD), Munich, Germany. 8. Department of Biochemistry, Yong Loo Lin School of Medicine, National University of Singapore, Singapore. 9. Department of Neurology, University Hospital, LMU Munich, Munich, Germany. 10. Institute of Neuroradiology, University Hospital, LMU Munich, Munich, Germany. 11. Institute of Laboratory Medicine, University Hospital, LMU Munich, Munich, Germany. 12. Institute of Clinical Pharmacology and Toxicology, University Medical Center Hamburg-Eppendorf, Hamburg, Germany. 13. German Centre for Cardiovascular Research (DZHK), Partner Site Hamburg / Kiel / Lübeck, Hamburg, Germany.
Abstract
OBJECTIVE: Early discrimination of patients with ischemic stroke (IS) from stroke mimics (SMs) poses a diagnostic challenge. The circulating metabolome might reflect pathophysiological events related to acute IS. Here, we investigated the utility of early metabolic changes for differentiating IS from SM. METHODS: We performed untargeted metabolomics on serum samples obtained from patients with IS (N = 508) and SM (N = 349; defined by absence of a diffusion weighted imaging [DWI] positive lesion on magnetic resonance imaging [MRI]) who presented to the hospital within 24 hours after symptom onset (median time from symptom onset to blood sampling = 3.3 hours; interquartile range [IQR] = 1.6-6.7 hours) and from neurologically normal controls (NCs; N = 112). We compared diagnostic groups in a discovery-validation approach by applying multivariable linear regression models, machine learning techniques, and propensity score matching. We further performed a targeted look-up of published metabolite sets. RESULTS: Levels of 41 metabolites were significantly associated with IS compared to NCs. The top metabolites showing the highest value in separating IS from SMs were asymmetrical and symmetrical dimethylarginine, pregnenolone sulfate, and adenosine. Together, these 4 metabolites differentiated patients with IS from SMs with an area under the curve (AUC) of 0.90 in the replication sample, which was superior to multimodal cranial computed tomography (CT; AUC = 0.80) obtained for routine diagnostics. They were further superior to previously published metabolite sets detected in our samples. All 4 metabolites returned to control levels by day 90. INTERPRETATION: A set of 4 metabolites with known biological effects relevant to stroke pathophysiology shows unprecedented utility to identify patients with IS upon hospital arrival, thus encouraging further investigation, including multicenter studies. ANN NEUROL 2020;88:736-746.
OBJECTIVE: Early discrimination of patients with ischemic stroke (IS) from stroke mimics (SMs) poses a diagnostic challenge. The circulating metabolome might reflect pathophysiological events related to acute IS. Here, we investigated the utility of early metabolic changes for differentiating IS from SM. METHODS: We performed untargeted metabolomics on serum samples obtained from patients with IS (N = 508) and SM (N = 349; defined by absence of a diffusion weighted imaging [DWI] positive lesion on magnetic resonance imaging [MRI]) who presented to the hospital within 24 hours after symptom onset (median time from symptom onset to blood sampling = 3.3 hours; interquartile range [IQR] = 1.6-6.7 hours) and from neurologically normal controls (NCs; N = 112). We compared diagnostic groups in a discovery-validation approach by applying multivariable linear regression models, machine learning techniques, and propensity score matching. We further performed a targeted look-up of published metabolite sets. RESULTS: Levels of 41 metabolites were significantly associated with IS compared to NCs. The top metabolites showing the highest value in separating IS from SMs were asymmetrical and symmetrical dimethylarginine, pregnenolone sulfate, and adenosine. Together, these 4 metabolites differentiated patients with IS from SMs with an area under the curve (AUC) of 0.90 in the replication sample, which was superior to multimodal cranial computed tomography (CT; AUC = 0.80) obtained for routine diagnostics. They were further superior to previously published metabolite sets detected in our samples. All 4 metabolites returned to control levels by day 90. INTERPRETATION: A set of 4 metabolites with known biological effects relevant to stroke pathophysiology shows unprecedented utility to identify patients with IS upon hospital arrival, thus encouraging further investigation, including multicenter studies. ANN NEUROL 2020;88:736-746.