Abstract
Background and Purpose
The cause of ischemic stroke remains unclear, or cryptogenic, in as many as 35% of stroke patients. Not knowing the cause of stroke restricts optimal implementation of prevention therapy and limits stroke research. We demonstrate how gene expression profiles in blood can be used in conjunction with a measure of infarct location on neuroimaging to predict a probable cause in cryptogenic stroke.
Methods
The cause of cryptogenic stroke was predicted using previously described profiles of differentially expressed genes characteristic of patients with cardioembolic, arterial and lacunar stroke. RNA was isolated from peripheral blood of 131 cryptogenic strokes and compared to profiles derived from 149 strokes of known cause. Each sample was run on Affymetrix U133 Plus2.0 microarrays. Cause of cryptogenic stroke was predicted using gene expression in blood and infarct location.
Results
Cryptogenic strokes were predicted to be 58% cardioembolic, 18% arterial, 12% lacunar and 12% unclear etiology. Cryptogenic stroke of predicted cardioembolic etiology had more prior myocardial infarction and higher CHA2DS2-VASc scores compared to stroke of predicted arterial etiology. Predicted lacunar strokes had higher systolic and diastolic blood pressures and lower NIHSS compared to predicted arterial and cardioembolic strokes. Cryptogenic strokes of unclear predicted etiology were less likely to have a prior TIA or ischemic stroke.
Conclusions
Gene expression in conjunction with a measure of infarct location can predict a probable cause in cryptogenic strokes. Predicted groups require further evaluation to determine whether relevant clinical, imaging, or therapeutic differences exist for each group.
Keywords: Ischemic stroke, cryptogenic stroke, gene expression, diagnosis
Introduction
The cause of stroke remains unclear, or cryptogenic, in as many as 35% of ischemic strokes despite careful clinical evaluation and diagnostic imaging 1, 2. This corresponds to over 280,000 cryptogenic strokes per year (32 per hour) in the US alone. Knowing the cause of stroke is essential to optimally implement prevention therapy. Given the two year mortality of cryptogenic strokes is 39%, there remains a significant need to prevent cardiovascular disease in this subtype of stroke 3. Accurate determination of stroke cause is also an important component in patient prognosis and in studies of stroke epidemiology, genetics and therapeutics. Thus, additional approaches to determine the cause of cryptogenic strokes are required.
When predicting the cause of cryptogenic stroke, stratification by infarct location can aid prediction. Cryptogenic small deep infarcts (SDI) occurring in the regions of the penetrating arteries require lacunar, arterial and cardiac disease all to be considered as a potential cause of stroke 4. However, in cryptogenic non-SDI only arterial and cardiac disease need to be considered as a potential cause of stroke. Other less common causes may contribute to either of these groups. Thus, infarct location can reduce the number of potential causes of stroke that need to be considered in the non-SDI group, which simplifies prediction models of cryptogenic stroke cause.
We recently described a gene expression profile to distinguish cardioembolic from arterial stroke, and a separate gene expression profile to distinguish lacunar from nonlacunar stroke 4, 5. In patients with cryptogenic stroke, these profiles may predict a probable cause to guide diagnostic and therapeutic studies. The profiles are based on differential inflammatory and prothrombotic states present in subtypes of stroke. In this study, we describe the integration of the two gene expression profiles with a measure of infarct location on neuroimaging to predict the probable cause of cryptogenic stroke.
Methods
1. Study Patients
Patients with acute ischemic stroke were enrolled from the University of California Davis and as part of the CLEAR trial at the University of Cincinnati as previously described 6 (NCT00250991 at Clinical-Trials.gov). The institutional review board at each site approved the study protocol and written informed consent was obtained from each patient. Patients with ischemic stroke were >18 years of age, had acute ischemia on neuroimaging and new neurological deficits persisting >24 hours. All patients had standardized clinical evaluations. Blood samples were drawn into PAXgene tubes (PreAnalytiX, Hilden, Germany) within 72 hours of stroke onset. A total of 131 cryptogenic stroke samples were analyzed using gene expression profiles derived from 149 strokes of known cause (79 cardioembolic, 40 arterial, 30 lacunar).
The cause of stroke was determined by 2 board-certified stroke neurologists based on medical history, blood tests, neuroimaging, Doppler, vascular angiography, electrocardiogram, echocardiogram and 24–48 hour cardiac monitoring. Strokes of arterial, cardioembolic and lacunar etiology that were used to derive the expression profiles have been previously described 4, 5. In brief, arterial strokes were patients with stenosis >50% in an intracranial or extracranial vessel referable to the infarct without other cause of stroke. Cardioembolic stokes included patients with atrial fibrillation, acute myocardial infarction, valvular heart disease, or marked ventricular hypokinesis without other cause of stroke. Patients with atrial myxoma or endocarditis were excluded. Lacunar strokes had a lacunar syndrome with a corresponding infarct <15mm in largest diameter in a region of the penetrating arteries without other cause of stroke. Cryptogenic strokes were patients with no determined etiology despite extensive evaluation, or patients with two or more potential causes of stroke so that a final diagnosis was not possible. Strokes with incomplete evaluation were excluded from study. Cryptogenic strokes were divided based on infarct location into two groups, SDI and non-SDI (Figure 1). Infarct location was identified by restricted diffusion on MRI or hypodensity on CT. Subjects without imaging evidence of infarction were excluded from study. Cryptogenic SDI had strokes in the regions of the penetrating arteries with infarcts >15mm in diameter, and/or two or more potential causes of stroke. Regions of the penetrating arteries included the basal ganglia, internal capsule, thalamus, corona radiata, and pons. Cryptogenic non-SDIs were strokes not in the regions of the penetrating arteries (mainly cortical), where lacunar stroke would not be considered a cause of stroke.
Figure 1.
Flow diagram for the prediction of cryptogenic stroke by infarct location and gene expression profiles. Cryptogenic strokes were divided into small deep infarcts (SDI) and non-SDI by location of infarct on neuroimaging. Cryptogenic SDI were predicted to be lacunar or non-lacunar stroke using a 41 gene profile. SDI of predicted non-lacunar stroke were then predicted to be arterial or cardioembolic stroke using a 40 gene profile. Cryptogenic non-SDI were predicted to be arterial or cardioembolic stroke using the 40 gene profile.
2. Sample Processing
Blood was collected by venipuncture into PAXgene tubes (PreAnalytiX, Germany). Samples were stored frozen and then processed at the same time to minimize technical variation. Total RNA was isolated according to protocol (PAXgene blood RNA kit; Pre-AnalytiX) and analyzed for quality using Agilent 2100 Bioanalyzer and quantity by Nano-drop. Reverse transcription, amplification, and sample labeling were carried out using Nugen’s Ovation Whole Blood reagents (Nugen Technologies, San Carlos, CA). Each RNA sample was hybridized to Affymetrix Genome U133 Plus 2 GeneChips and scanned according to protocol (Affymetrix Santa Clara, CA). Raw expression values were pre-processed using robust multichip averaging (RMA), mean-centering standardization and log2 transformation for prediction analyses (Partek Genomics Suite 6.4, Partek Inc., St. Louis, MO).
3. Gene Profiles
The profiles used to predict cause in cryptogenic stroke have been described previously. A 41 gene list was identified to distinguish patients with lacunar stroke from non-lacunar stroke 5 (Supplementary Table 1). In addition, a 40 gene list was identified to distinguish patients with cardioembolic stroke from patients with arterial stroke 4 (Figure 1) (Supplementary Table 2). These gene profiles were derived from patients with known cardioembolic, arterial and lacunar causes of stroke.
4. Prediction of Cryptogenic Stroke
Prediction analyses were performed using linear discriminant analysis (Partek Genomics Suite 6.4) and nearest shrunken centroid (Prediction Analysis of Microarrays, PAM) algorithms and principal components analysis (Partek Genomics Suite 6.4) (Figure 2)7. Patients with known cause of stroke were used to develop prediction models to distinguish cardioembolic from arterial causes of stroke 4 and to distinguish lacunar from nonlacunar causes of stroke 5. These models were applied to groups of cryptogenic stroke patients to predict the cause of stroke (Figure 1). The prediction models assign a probability of belonging to a type of stroke based on the pattern of gene expression. For example, cryptogenic strokes with a pattern of gene expression similar to patients with known cardioembolic stroke are predicted to be cardioembolic stroke. This prediction is performed to determine which cryptogenic strokes are most likely cardioembolic, arterial or lacunar.
Figure 2.
Principal components analysis (PCA) of cryptogenic strokes using the genes predictive of stroke cause. A. PCA of non-SDI cryptogenic strokes predicted to be arterial or cardioembolic stroke using the 40 gene profile. B. PCA of SDI cryptogenic strokes predicted to be lacunar or non-lacunar stroke using the 41 gene profile. C. PCA of the cryptogenic SDI predicted to be non-lacunar stroke in B, predicted to be arterial or cardioembolic stroke using the 40 gene profile.
Prediction of cryptogenic strokes was performed using gene expression in conjunction with infarct location. Cryptogenic SDI in regions of the penetrating arteries were first predicted to be of either lacunar or nonlacunar etiology using the 41 gene profile (Figure 1) (Supplementary Table 1). Cryptogenic SDIs of predicted nonlacunar etiology were then predicted to be of cardioembolic or arterial etiology using the 40 gene profile. Cryptogenic non-SDI (mostly cortical infarcts) were predicted to be of cardioembolic or arterial etiology using the 40 gene profile (Supplementary Table 2). The 41 gene profile was not used to predict this group because lacunar stroke is not a consideration in patients with non-SDI. A predicted diagnosis of stroke cause required an average predicted probability of >75%. Subjects not meeting these criteria remained of unclear cause.
5. Statistical analyses
Differences in demographic and clinical data between predicted stroke groups were analyzed using Chi-square or Fisher’s exact test, Kruskal Wallis, or Analysis of Variance as appropriate (Stata 10.1, College Station, TX, USA). Data are presented as mean ± standard deviation (SD) for continuous variables and median with interquartile range (IQR) for ordinal variables.
Results
Cryptogenic Strokes
The demographic and clinical features of the 131 cryptogenic stroke patients are summarized in Table 1. The average age was 63.9 years (SD 14.8) and 54.2% were male. Cryptogenic strokes were of diverse race and ethnicity, with 74% white, 16% black, 6.1% Hispanic, and 3.9% of other race. Hypertension was present in 62.6%, hyperlipidemia in 25.2%, diabetes in 18.3%. 21.4% had prior stroke or TIA, 15.3% had prior myocardial infarction, and 3.1% had atrial fibrillation. The median NIHSS on admission was 10.6 (IQR 5.5,15.6). There were 32 (24.4%) SDI and 99 (75.6%) non-SDI infarcts. The demographics of the cryptogenic stroke patients were similar to strokes of known etiology used to derive the gene expression profiles. The average age for strokes of known etiology was 67.7 (SD 10.4), 59.1% were male, 73.2% had hypertension, 31.5% had hyperlipidemia and 30.2% had diabetes (Supplementary Table 3)
Table 1.
Characteristics of the 131 cryptogenic stroke subjects studied
| Characteristic | Value |
|---|---|
| Age years (SD) | 63.9 (14.8) |
| Male n(%) | 71 (54.2%) |
| Hypertension n(%) | 82 (62.6%) |
| Systolic BP mmHg (SD) | 160.1 (36.0) |
| Diastolic BP mmHg (SD) | 86.8 (18.7) |
| Diabetes n(%) | 24 (18.3%) |
| Body mass index kg/m2 (SD) | 28.7 (4.5) |
| Hyperlipidemia n(%) | 33 (25.2%) |
| Current Smoking n(%) | 44 (33.6%) |
| Atrial fibrillation n(%) | 4 (3.1%) |
| Prior Myocardial infarction n(%) | 20 (15.3%) |
| Congestive heart failure n(%) | 6 (4.6%) |
| Prior Stroke or TIA n(%) | 28 (21.4%) |
| NIHSS Admission (IQR) | 10.6 (6–16) |
Cryptogenic Stroke Predicted by Gene Expression and Infarct Location
Of the 32 SDI cryptogenic strokes 11 (34.4%) were predicted to be cardioembolic, 4 (12.5%) to be arterial, 15 (46.9%) to be lacunar, and 2 (6.3%) remained of unclear etiology. Of the 99 non-SDI cryptogenic strokes, 65 (65.7%) were predicted to be cardioembolic, 20 (20.2%) to be arterial and 14 (14.1%) remained of unclear etiology. Thus, when combined, the 131 cryptogenic strokes resulted in 76 (58.0%) subjects predicted to be cardioembolic, 24 (18.3%) to be arterial, and 15 (11.5%) to be lacunar (Figure 1). There remained 16 (12.2%) subjects in which the diagnostic probability was insufficient to assign a predicted class. Separation of cryptogenic stroke into predicted causes by gene expression profiles is shown by principal components analysis in Figure 2 (Partek Genomics Suite 6.4)7.
The characteristics of cryptogenic strokes divided by predicted etiology is shown in Table 2. Groups were similar in gender, hypertension, diabetes and hyperlipidemia. Cryptogenic strokes of predicted cardioembolic etiology had more prior myocardial infarction and higher CHA2DS2-VASc scores compared to cryptogenic strokes of predicted arterial etiology. Cryptogenic strokes of predicted lacunar etiology had higher systolic and diastolic blood pressures and a lower NIHSS compared to cryptogenic strokes predicted to be arterial or cardioembolic. Cryptogenic strokes that remained of unclear etiology had less prior history of stroke or TIA.
Table 2.
Characteristics of cryptogenic strokes divided by their predicted etiology based on gene expression profiles and infarct location. Cryptogenic strokes were predicted to be of cardioembolic, arterial, or lacunar etiology. Strokes of unclear predicted cause were subjects where the diagnostic probability was insufficient to assign a predicted class
| Cardioembolic (n=76) | Arterial (n=24) | Lacunar (n=15) | Unclear (n=16) | p- value | |
|---|---|---|---|---|---|
| Age years (SD) | 65.2 (16.5) | 61.1 (11.5) | 62.9 (13.7) | 61.1 (11.5) | <0.05 |
| Male n(%) | 35 (46.1%) | 15 (62.5%) | 9 (60.0%) | 12 (75.0%) | 0.13 |
| Hypertension n(%) | 46 (60.6%) | 17 (70.8%) | 12 (80.0%) | 7 (43.8%) | 0.1 |
| Systolic BP mmHg (SD) | 159.4 (37.9) | 155.0 (38.1) | 175.9 (37.3) | 155.6 (17.6) | <0.05 |
| Diastolic BP mmHg (SD) | 85.5 (16.9) | 81.0 (21.2) | 97.9 (21.3) | 90.1 (17.6) | <0.05 |
| Diabetes n(%) | 12 (15.8%) | 6 (25.0%) | 5 (33.3%) | 1 (6.3%) | 0.15 |
| Body mass index kg/m2 (SD) | 28.5 (3.9) | 29.0 (2.7) | 28.4 (5.1) | 31.6 (9.3) | 0.61 |
| Hyperlipidemia n(%) | 19 (25.0%) | 4 (16.7%) | 6 (40%) | 4 (25.0%) | 0.35 |
| Current Smoking n(%) | 21 (27.6%) | 10 (41.7%) | 6 (40.0%) | 7 (43.8%) | 0.58 |
| Atrial fibrillation n(%) | 3 (3.9%) | 0 | 1 (6.7%) | 0 | 0.02 |
| Prior myocardial infarction n(%) | 17 (22.4%) | 0 | 0 | 3 (18.8%) | 0.02 |
| Congestive heart failure n(%) | 6 (7.9%) | 0 | 0 | 0 | 0.42 |
| CHA2DS2-VASc (IQR) | 3.1 (2–4) | 2.0 (1–2.7) | 2.8 (1.2–3.7) | 1.6 (0–3) | 0.05 |
| Prior Stroke or TIA n(%) | 21 (27.6%) | 3 (12.5%) | 4 (26.7%) | 0 | 0.04 |
| NIHSS Admission (IQR) | 9.6 (6–11.5) | 12.1 (6–19) | 3.3 (2–3.5) | 16.6 (12–23) | <0.05 |
Discussion
We describe the use of RNA expression profiles in blood in conjunction with infarct location to identify probable cardioembolic, arterial and lacunar causes of cryptogenic stroke. The predicted causes of cryptogenic stroke require further study to determine whether relevant epidemiologic or therapeutic differences exist for each group. The described approach to determining a probable cause of cryptogenic stroke represents an important step to better classify a patient group that accounts for as many as one third of ischemic strokes.
Cardioembolic stroke was the predicted cause in 58% of cryptogenic strokes studied. This is consistent with previous studies indicating cardioembolism is a contributor to cryptogenic stroke 8–10. Indeed, paroxysmal atrial fibrillation not detected on initial investigations can later be identified in 10–25% of cryptogenic strokes with prolonged cardiac monitoring 8, 11. In our study, a higher CHA2DS2-VASc score in predicted cardioembolic subjects may support the presence of undetected paroxysmal atrial fibrillation. This requires further evaluation by prolonged cardiac recording. Patent foramen ovale (PFO) may also contribute to cryptogenic stroke, though this remains controversial 12–14. The prevalence of PFO is increased in cryptogenic strokes compared to controls 15, 16, and trials of PFO closure in cryptogenic stroke are ongoing 17. Whether cryptogenic strokes of predicted cardioembolic etiology may derive greater benefit from PFO closure may be of interest for future evaluation. Prior myocardial infarction was also identified to be more common in cryptogenic strokes of predicted cardioembolic etiology. Though the significance of this remains unclear, a prior myocardial infarction in cryptogenic stroke may be an indicator of underlying cardiac disease associated with stroke.
Arterial disease was a predicted cause of cryptogenic stroke in 18.3% of subjects. Prior studies also suggest that arterial disease likely contributes to cryptogenic stroke, including arterial stenosis <50%, plaque ulceration, intracranial atherosclerosis and aortic plaque 10, 18, 19. Indeed, arterial stenosis <50% and the presence of arterial plaque are both considered in the Causative Classification for Ischemic Stroke (CSS) and the A-S-C-O classification systems as features that increase the probability of arterial stroke 20, 21. Ulcerated plaques missed by current investigations may also account for some of the predicted arterial cryptogenic strokes, particularly given that two thirds of ulcerated carotid arteries occur with stenoses of 0–49% 22. Further evaluation to determine whether such features are more common in cryptogenic strokes of predicted arterial etiology is required.
Lacunar stroke was a predicted cause of cryptogenic stroke in 11.5% of subjects. Lacunar stroke differs from arterial and cardioembolic stroke in that the source of stroke is inferred by infarct size and location rather than a directly identified source of stroke. Thus, lacunar stroke is only a consideration when an infarct is located in the region of the penetrating arteries. These small deep infarcts are cryptogenic when an arterial and/or cardioembolic source of stroke is present or when the infarct is larger than 15mm in diameter. The prediction of cause by gene expression suggests that some small deep infarcts greater than 15mm in size, or with coincidental arterial or cardiac source of stroke, are lacunar strokes. Evaluating these predicted causes will require follow up. Subjects of predicted lacunar etiology would be expected to be more likely to have recurrent lacunar infarcts, whereas SDIs of predicted arterial or cardioembolic etiology would be expected to have recurrent infarcts in regions not supplied by deep penetrating arteries.
In this study 76% of cryptogenic strokes were predicted to be of either arterial or cardioembolic etiology. This is consistent with previous epidemiological studies indicating cryptogenic strokes have a long term prognosis similar to that of arterial and cardioembolic stroke. The 2 year survival of cryptogenic stroke is 61%, which is closer to the survival of cardioembolic stroke at 55% and arterial stroke at 58% compared to lacunar stroke at 85% 3. Additionally, cryptogenic strokes are more likely to have recurrent arterial or cardioembolic strokes 23. Finally outcome in cryptogenic stroke is closer to that of cardioembolic and arterial stroke than lacunar stroke. At 90 days 35.8% of cryptogenic strokes have an m-Rankin score >3 24, compared to 56.8% of cardioembolic strokes, 32.4% of arterial strokes and 4.2% of lacunar strokes 24.
The CSS and A-S-C-O classification systems demonstrate how additional clinical and imaging features can be used to further characterize stroke subtype. Likewise, genetic data may also provide additional information to characterize stroke subtype. We demonstrate how RNA expression profiles can predict a likely cause in cryptogenic stroke. As with other diagnostic tests, a gene expression profile could be developed into a clinical assay similar to those developed for breast cancer and coronary artery disease 25, 26. The RNA results would provide an indication of whether a patient is likely of arterial, cardioembolic, or lacunar etiology, which could be interpreted along with other diagnostic information to determine stroke cause.
This discovery type study has limitations. Although echocardiography was performed in all subjects, transesophageal echocardiography was not. Contrast enhanced MRA was rarely performed in subjects, though MRA or CTA was performed on nearly all subjects. Evaluation for coagulopathy was performed in patients <50 years of age or in cases where physicians deemed appropriate. Studies with long-term follow up are needed to document recurrent stroke with potentially identifiable cause and to monitor for arterial, cardioembolic or small vessel disease. Additionally, the relationship between gene expression prediction of cryptogenic stroke and other features associated with stroke cause require further study, including imaging criteria and features specified in the CSS and A-S-C-O stroke classification systems. The profiles used to predict the cause of cryptogenic stroke were derived from a relatively small sample size. Though vascular risk factors, gender, race and ethnicity were not found to be significantly different between groups, further development and evaluation of RNA predictors of stroke cause in larger cohorts is required to ensure adequate representation of a stroke population.
Conclusions
The high rates of recurrent stroke and mortality in cryptogenic stroke support the need to develop methods to diagnose this large group of patients. Identifying a cause of stroke is critical to implement prevention therapy and reduce stroke disability and mortality. We demonstrate a probable cause of cryptogenic stroke can be predicted based on gene expression profile in conjunction with a measure of infarct location. Further study of predicted groups is required to determine whether relevant clinical, imaging or therapeutic differences exist for each group.
Supplementary Material
Acknowledgments
We thank the investigators of the SPOTRIAS Stroke Network involved in the CLEAR trial for supplying blood samples for analysis. We appreciate the support of the M.I.N.D. Institute, the Genomics and Expression Resource at the M.I.N.D. Institute, and the UCD Department of Neurology.
Sources of Funding
This study was supported by the NIH (NS056302, FRS) and the American Heart Association (FRS). GCJ receives support from the Canadian Institutes of Health Research (CIHR) and the American Heart Association.
Footnotes
Disclosures
None.
References
- 1.Adams HP, Jr, Bendixen BH, Kappelle LJ, Biller J, Love BB, Gordon DL, et al. Classification of subtype of acute ischemic stroke. Definitions for use in a multicenter clinical trial. Toast. Trial of org 10172 in acute stroke treatment. Stroke. 1993;24:35–41. doi: 10.1161/01.str.24.1.35. [DOI] [PubMed] [Google Scholar]
- 2.Ionita CC, Xavier AR, Kirmani JF, Dash S, Divani AA, Qureshi AI. What proportion of stroke is not explained by classic risk factors? Prev Cardiol. 2005;8:41–46. doi: 10.1111/j.1520-037x.2005.3143.x. [DOI] [PubMed] [Google Scholar]
- 3.Kolominsky-Rabas PL, Weber M, Gefeller O, Neundoerfer B, Heuschmann PU. Epidemiology of ischemic stroke subtypes according to toast criteria: Incidence, recurrence, and long-term survival in ischemic stroke subtypes: A population-based study. Stroke. 2001;32:2735–2740. doi: 10.1161/hs1201.100209. [DOI] [PubMed] [Google Scholar]
- 4.Jickling GC, Xu H, Stamova B, Ander BP, Zhan X, Tian Y, et al. Signatures of cardioembolic and large-vessel ischemic stroke. Ann Neurol. 2010;68:681–692. doi: 10.1002/ana.22187. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Jickling GC, Stamova B, Ander BP, Zhan X, Tian Y, Liu D, et al. Profiles of lacunar and nonlacunar stroke. Ann Neurol. 2011;70:477–485. doi: 10.1002/ana.22497. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Pancioli AM, Broderick J, Brott T, Tomsick T, Khoury J, Bean J, et al. The combined approach to lysis utilizing eptifibatide and rt-pa in acute ischemic stroke: The clear stroke trial. Stroke. 2008;39:3268–3276. doi: 10.1161/STROKEAHA.108.517656. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Jolliffe IT. Principal component analysis. New York: Springer; 2002. [Google Scholar]
- 8.Stollberger C, Finsterer J. Detection of paroxysmal atrial fibrillation and patent foramen ovale in cryptogenic stroke. Eur J Neurol. 2009;16:160–161. doi: 10.1111/j.1468-1331.2008.02410.x. [DOI] [PubMed] [Google Scholar]
- 9.Elijovich L, Josephson SA, Fung GL, Smith WS. Intermittent atrial fibrillation may account for a large proportion of otherwise cryptogenic stroke: A study of 30-day cardiac event monitors. J Stroke Cerebrovasc Dis. 2009;18:185–189. doi: 10.1016/j.jstrokecerebrovasdis.2008.09.005. [DOI] [PubMed] [Google Scholar]
- 10.Amarenco P. Underlying pathology of stroke of unknown cause (cryptogenic stroke) Cerebrovasc Dis. 2009;27 (Suppl 1):97–103. doi: 10.1159/000200446. [DOI] [PubMed] [Google Scholar]
- 11.Tayal AH, Tian M, Kelly KM, Jones SC, Wright DG, Singh D, et al. Atrial fibrillation detected by mobile cardiac outpatient telemetry in cryptogenic tia or stroke. Neurology. 2008;71:1696–1701. doi: 10.1212/01.wnl.0000325059.86313.31. [DOI] [PubMed] [Google Scholar]
- 12.Alsheikh-Ali AA, Thaler DE, Kent DM. Patent foramen ovale in cryptogenic stroke: Incidental or pathogenic? Stroke. 2009;40:2349–2355. doi: 10.1161/STROKEAHA.109.547828. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Saver JL. Cryptogenic stroke in patients with patent foramen ovale. Curr Atheroscler Rep. 2007;9:319–325. doi: 10.1007/s11883-007-0039-y. [DOI] [PubMed] [Google Scholar]
- 14.Handke M, Harloff A, Olschewski M, Hetzel A, Geibel A. Patent foramen ovale and cryptogenic stroke in older patients. N Engl J Med. 2007;357:2262–2268. doi: 10.1056/NEJMoa071422. [DOI] [PubMed] [Google Scholar]
- 15.Overell JR, Bone I, Lees KR. Interatrial septal abnormalities and stroke: A meta-analysis of case-control studies. Neurology. 2000;55:1172–1179. doi: 10.1212/wnl.55.8.1172. [DOI] [PubMed] [Google Scholar]
- 16.Cabanes L, Mas JL, Cohen A, Amarenco P, Cabanes PA, Oubary P, et al. Atrial septal aneurysm and patent foramen ovale as risk factors for cryptogenic stroke in patients less than 55 years of age. A study using transesophageal echocardiography. Stroke. 1993;24:1865–1873. doi: 10.1161/01.str.24.12.1865. [DOI] [PubMed] [Google Scholar]
- 17.Furlan AJ, Reisman M, Massaro J, Mauri L, Adams H, Albers GW, et al. Study design of the closure i trial: A prospective, multicenter, randomized, controlled trial to evaluate the safety and efficacy of the starflex septal closure system versus best medical therapy in patients with stroke or transient ischemic attack due to presumed paradoxical embolism through a patent foramen ovale. Stroke. 2010;41:2872–2883. doi: 10.1161/STROKEAHA.110.593376. [DOI] [PubMed] [Google Scholar]
- 18.Harloff A, Dudler P, Frydrychowicz A, Strecker C, Stroh AL, Geibel A, et al. Reliability of aortic mri at 3 tesla in patients with acute cryptogenic stroke. J Neurol Neurosurg Psychiatry. 2008;79:540–546. doi: 10.1136/jnnp.2007.125211. [DOI] [PubMed] [Google Scholar]
- 19.Kitagawa K. Aortogenic embolism is a possible mechanism of cryptogenic stroke. Ann Neurol. 2004;55:598–599. doi: 10.1002/ana.20073. [DOI] [PubMed] [Google Scholar]
- 20.Amarenco P, Bogousslavsky J, Caplan LR, Donnan GA, Hennerici MG. New approach to stroke subtyping: The a-s-c-o (phenotypic) classification of stroke. Cerebrovasc Dis. 2009;27:502–508. doi: 10.1159/000210433. [DOI] [PubMed] [Google Scholar]
- 21.Ay H, Benner T, Arsava EM, Furie KL, Singhal AB, Jensen MB, et al. A computerized algorithm for etiologic classification of ischemic stroke: The causative classification of stroke system. Stroke. 2007;38:2979–2984. doi: 10.1161/STROKEAHA.107.490896. [DOI] [PubMed] [Google Scholar]
- 22.Homburg PJ, Rozie S, van Gils MJ, van den Bouwhuijsen QJ, Niessen WJ, Dippel DW, et al. Association between carotid artery plaque ulceration and plaque composition evaluated with multidetector ct angiography. Stroke. 2011;42:367–372. doi: 10.1161/STROKEAHA.110.597369. [DOI] [PubMed] [Google Scholar]
- 23.Bang OY, Lee PH, Joo SY, Lee JS, Joo IS, Huh K. Frequency and mechanisms of stroke recurrence after cryptogenic stroke. Ann Neurol. 2003;54:227–234. doi: 10.1002/ana.10644. [DOI] [PubMed] [Google Scholar]
- 24.Petty GW, Brown RD, Jr, Whisnant JP, Sicks JD, O’Fallon WM, Wiebers DO. Ischemic stroke subtypes : A population-based study of functional outcome, survival, and recurrence. Stroke. 2000;31:1062–1068. doi: 10.1161/01.str.31.5.1062. [DOI] [PubMed] [Google Scholar]
- 25.Sotiriou C, Pusztai L. Gene-expression signatures in breast cancer. N Engl J Med. 2009;360:790–800. doi: 10.1056/NEJMra0801289. [DOI] [PubMed] [Google Scholar]
- 26.Rosenberg S, Elashoff MR, Beineke P, Daniels SE, Wingrove JA, Tingley WG, et al. Multicenter validation of the diagnostic accuracy of a blood-based gene expression test for assessing obstructive coronary artery disease in nondiabetic patients. Ann Intern Med. 2010;153:425–434. doi: 10.7326/0003-4819-153-7-201010050-00005. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.