Plasma lipidomic profiling reveals six candidate biomarkers for the prediction of incident stroke in patients with hypertension.

Abstract

The burden of stroke in patients with hypertension is very high, and its prediction is critical. We aimed to use plasma lipidomics profiling to identify lipid biomarkers for predicting incident stroke in patients with hypertension. This was a nested case-control study. Baseline plasma samples were collected from 30 hypertensive patients with newly developed stroke, 30 matched patients with hypertension, 30 matched patients at high risk of stroke, and 30 matched healthy controls. Lipidomics analysis was performed by ultrahigh-performance liquid chromatography-tandem mass spectrometry, and differential lipid metabolites were screened using multivariate and univariate statistical methods. Machine learning methods (least absolute shrinkage and selection operator, random forest) were used to identify candidate biomarkers for predicting stroke in patients with hypertension. Co-expression network analysis revealed that the key molecular alterations of the lipid network in stroke implicate glycerophospholipid metabolism and choline metabolism. Six lipid metabolites were identified as candidate biomarkers by multivariate statistical and machine learning methods, namely phosphatidyl choline(40:3p)(rep), cholesteryl ester(20:5), monoglyceride(29:5), triglyceride(18:0p/18:1/18:1), triglyceride(18:1/18:2/21:0) and coenzyme(q9). The combination of these six lipid biomarkers exhibited good diagnostic and predictive ability, as it could indicate a risk of stroke at an early stage in patients with hypertension (area under the curve = 0.870; 95% confidence interval: 0.783-0.957). We determined lipidomic signatures associated with future stroke development and identified new lipid biomarkers for predicting stroke in patients with hypertension. The biomarkers have translational potential and thus may serve as blood-based biomarkers for predicting hypertensive stroke.