Objective We tested the hypothesis that the risk of intracranial hemorrhage (ICH) in patients with cardioembolic ischemic stroke who are treated with oral anticoagulants (OAs) can be predicted by evaluating surrogate markers of hemorrhagic-prone cerebral angiopathies using a baseline MRI. Methods Patients were participants in a multicenter and prospective observational study. They were older than 64 years, had a recent cardioembolic ischemic stroke, and were new users of OAs. They underwent a baseline MRI analysis to evaluate microbleeds, white matter hyperintensities, and cortical superficial siderosis. We collected demographic variables, clinical characteristics, risk scores, and therapeutic data. The primary endpoint was ICH that occurred during follow-up. We performed bivariate and multivariate Cox regression analyses. Results We recruited 937 patients (aged 77.6 ± 6.5 years; 47.9% were men). Microbleeds were detected in 207 patients (22.5%), moderate/severe white matter hyperintensities in 419 (45.1%), and superficial siderosis in 28 patients (3%). After a mean follow-up of 23.1 ± 6.8 months, 18 patients (1.9%) experienced an ICH. In multivariable analysis, microbleeds (hazard ratio 2.7, 95% confidence interval [CI] 1.1–7, p = 0.034) and moderate/severe white matter hyperintensities (hazard ratio 5.7, 95% CI 1.6–20, p = 0.006) were associated with ICH (C index 0.76, 95% CI 0.66–0.85). Rate of ICH was highest in patients with both microbleed and moderate/severe WMH (3.76 per 100 patient-years, 95% CI 1.62–7.4). Conclusion Patients taking OAs who have advanced cerebral small vessel disease, evidenced by microbleeds and moderate to severe white matter hyperintensities, had an increased risk of ICH. Our results should help to determine the risk of prescribing OA for a patient with cardioembolic stroke. ClinicalTrials.gov identifier NCT02238470.
Background and Purpose—: Restenosis after carotid angioplasty (with or without stent) is associated with increased rate of stroke and death. Our aim was to determine risk and predictive factors related to carotid restenosis post carotid angioplasty and its association to recurrent cerebrovascular events. Methods—: All consecutive patients with carotid stenosis treated with angioplasty (n=1060) in a single University Hospital were included (from 2002 to 2013). Follow-up was done prospectively evaluating restenosis, ipsilateral stroke, or death. Restenosis was defined as a narrowing of ≥70% of a previously treated vessel evaluated by ultrasonography. Results—: Of the 1060 patients treated, 9.2% (97) of patients experienced restenosis during follow up (median 12 [9–32] months). Occurrence of restenosis was associated with ipsilateral stroke during follow-up ( P =0.049). After Cox regression analysis, hypertension (hazard ratio, 6.2 [1.9–19.9]; P =0.002), impaired vasoreactivity (hazard ratio, 1.7 [1.09–2.8]; P =0.019), and angioplasty without stent (hazard ratio, 2.9 [1.2–6.8]; P =0.012) were independent risk predictors of >70% restenosis. Conclusions—: Carotid restenosis after carotid angioplasty is associated with ipsilateral stroke occurrence. In our sample, hypertension, angioplasty without stent, and impaired vasoreactivity identify patients at high risk of restenosis and could help to select patients for follow-up ultrasonography imaging.
Single-cell RNA sequencing (scRNA-seq) quantifies gene expression for individual cells in a sample, which allows distinct cell-type populations to be identified and characterized. An important step in many scRNA-seq analysis pipelines is the annotation of cells into known cell-types. While this can be achieved using experimental techniques, such as fluorescence-activated cell sorting, these approaches are impractical for large numbers of cells. This motivates the development of data-driven cell-type annotation methods. We find limitations with current approaches due to the reliance on known marker genes or from overfitting because of systematic differences between studies or batch effects. Here, we present a statistical approach that leverages public datasets to combine information across thousands of genes, uses a latent variable model to define cell-type-specific barcodes and account for batch effect variation, and probabilistically annotates cell-type identity. The barcoding approach also provides a new way to discover marker genes. Using a range of datasets, including those generated to represent imperfect real-world reference data, we demonstrate that our approach substantially outperforms current reference-based methods, in particular when predicting across studies. Our approach also demonstrates that current approaches based on unsupervised clustering lead to false discoveries related to novel cell-types.