American Heart Association

pathogenesis

Machine Learning as a Tool for Etiological Investigation in Stroke Medicine

Aurora Semerano, MD
@semerano_aurora

Kamel H, Navi BB, Parikh NS, Merkler AE, Okin PM, Devereux RB, Weinsaft JW, Kim J, Cheung JW, Kim LK, et al. Machine Learning Prediction of Stroke Mechanism in Embolic Strokes of Undetermined Source. Stroke. 2020;51:e203–e210.

In 2014, when the concept of embolic stroke of undetermined source (ESUS) was proposed,1 confidence existed that ESUS could represent a single entity which would have benefitted from a unified treatment. However, after two randomized clinical trials did not show benefit of direct oral anticoagulation for secondary prevention of ESUS patients,2,3 it is now common opinion that these patients rather represent a heterogeneous population and are likely to benefit from tailored, personalized therapies. Today, ESUS represents a useful definition to identify patients deserving extended diagnostic workup, while prevention therapy for these patients remains elusive, and clinical stroke recurrence is still an issue. Both subgroup analyses from the above-mentioned clinical trials and new research studies have been developed or are ongoing, to better understand the pathophysiology of ESUS and help in patient selection.

In such a phenotypically heterogeneous population, one big effort is to identify patient subsets with a single or group of underlying mechanisms likely to respond to an established treatment. With this right purpose of uncover the “hidden structure” in a complex scenario, the recent study from Kamel et al.4 employs a machine learning approach. Firstly, a supervised machine-learning algorithm was developed to distinguish cardioembolic versus non-cardioembolic strokes in a population of 1083 patients with known stroke etiology, by entering data about demographics, comorbidities, vitals, laboratory results, and echocardiograms. After the learning process, the system finally resulted to distinguish cardioembolic from non-cardioembolic strokes with excellent accuracy (area under the curve, AUC=0.85).

Article Commentary: “Nonfocal Transient Neurological Attacks Are Associated With Cerebral Small Vessel Disease”

Richard Jackson, MD

Oudeman EA, Greving JP, Van den Berg-Vos RM, Biessels GJ, Bron EE, van Oustenbrugge R, et al. Nonfocal Transient Neurological Attacks Are Associated With Cerebral Small Vessel Disease. Stroke. 2019;50:3540–3544.

Oudeman et al. have published a paper on non-focal TIA’s which have been called transient neurological attacks (TNAs). As neurologists, we are always trained to localize first, but anyone who has taken stroke calls from the ER knows that not all the presentations and calls are localizable, such as those mentioned in this paper and described as “bilateral weakness, unsteadiness, or confusion.” In the introduction, the authors mention that these were historically thought to be secondary to hypoperfusion, but recent imaging have shown them to be associated with ischemia. Underlying risk factors are common to traditional TIA’s, such as smoking and hypertension, but not atherosclerotic disease or atrial fibrillation, which led to the current investigation for an association between small vessel disease and TNAs.

Effect of HRV on the Association Between Obstructive Sleep Apnea and Small Vessel Disease

Kristina Shkirkova, BSc
@KShkirkova

Del Brutto OH, Mera RM, Costa AF, Castillo PR. Effect of Heart Rate Variability on the Association Between the Apnea-Hypopnea Index and Cerebral Small Vessel Disease. Stroke. 2019;50:2486–2491.

Obstructive Sleep Apnea (OSA) is a form of sleep-disordered breathing that has been increasingly implicated in the pathogenesis of cerebral small vessel disease (cSVD). OSA is associated with recurrent episodes of hypoxia, altered cerebral autoregulation, and sympathetic overactivity, which may be contributing triggers for pathophysiology of cSVD. A recent study by Del Brutto et al. used nighttime Heart Rate Variability (HRV) as a measure of sympathetic upregulation to study the association between OSA and cSVD. HRV measures variation in the intervals between heartbeats and is used as a reflection of the balance between sympathetic and parasympathetic tone. Apnea-Hypopnea Index was used to access the degree of OSA and the total cSVD score was chosen to quantify cSVD burden. The study used data from the Atahualpa Project, which included elderly (age above 60) residents of the Atahualpa rural village on the coast of Ecuador. A total of 176 participants who underwent clinical assessment, magnetic resonance imaging (MRI), single-night polysomnography, and 24-hour Holter monitoring were selected for the analysis.

Among study participants, the mean age was 71.8, and 64% were women. The univariate analysis showed that daytime HRV below the 50th percentile was associated with female gender and lower mean percentage of O2 saturation. The nighttime HRV below the 50th percentile was associated with body mass index (BMI) higher than 30 kg/m2. In the generalized linear model analysis, with and without confounding variables, there was a significant association between the cSVD score and AHI (p=0.026). Furthermore, a negative association was observed between sCVD and nighttime HRV, but not daytime HRV (p=0.001). Interaction model analysis showed a significant interaction of nighttime HRV on the relationship between AHI and the cSVD score (P=0.001). The total effect between AHI and the cSVD score mediated by HRV was 30.8%. Additionally, contour plots showed the effect of nighttime HRV on the association between AHI and the cSVD score.

Clot Histology: A Possible Clue to the Etiology of Ischemic Stroke

Piyush Ojha, MBBS, MD, DM

Fitzgerald S, Dai D, Wang S, Douglas A, Kadirvel R, Layton KF, et al. Platelet-Rich Emboli in Cerebral Large Vessel Occlusion Are Associated With a Large Artery Atherosclerosis Source. Stroke. 2019;50:1907–1910.

Stroke accounts for approximately 10% of all deaths worldwide and leads to substantial long-term disability. The majority of the strokes are ischemic in origin. No identifiable cause is found in up to one-third of the patients after a standard evaluation, which limits the options for secondary stroke prevention. Mechanical thrombectomy has been found to be highly effective in patients with large vessel occlusions (LVO). In addition to the revascularisation, endovascular procedures have also created a unique opportunity to identify the likely stroke pathogenesis by providing thrombus material for further study. Emerging insights on various thrombus characteristics can not only provide valuable information that might be useful for guiding acute therapies, but also in optimizing secondary stroke prevention, as different components in the clot may respond to different pharmacological strategies.

Studies have tried to correlate thrombus histological composition and stroke pathogenesis. Sporns et al.1 observed that clots from a cardioembolic source had a higher proportion of fibrin/platelets and fewer red blood cells than noncardioembolic thrombi.

Atrial Thromboembolism: Looking Beyond Atrial Fibrillation

Andrea Morotti, MD

Kamel H, Bartz TM, Elkind MSV, Okin PM, Thacker EL, Patton KK, et al. Atrial Cardiopathy and the Risk of Ischemic Stroke in the CHS (Cardiovascular Health Study). Stroke. 2018

Many patients experiencing cryptogenic stroke with an embolic pattern lack a documented source of thromboembolism, even after an extensive diagnostic workup and prolonged follow-up. In particular, atrial fibrillation (AF) is detected in less than half of these patients. Increasing evidence suggested that atrial dysfunction may lead to thromboembolism independently of AF. In this large, multicenter prospective study, Dr. Kamel and colleagues explored the association between markers of atrial cardiopathy and the risk of incident ischemic stroke.

Resistant Atherosclerosis

Philip Chang, MD

Spence JD, Solo K. Resistant Atherosclerosis: The Need for Monitoring of Plaque Burden. Stroke. 2017

In this study, Spence and Solo demonstrated that measurement of LDL-C levels is likely inadequate to assess a patient’s response to statin therapy. In their database of 4512 patients with 2 measurements of LDL-C and 2 carotid duplex scans measuring total plaque area, they found that neither LDL-C levels nor change in LDL-C levels predicted carotid artery plaque burden or progression of plaque area. Interestingly, they found that in the 6% of patients with low LDL-C levels (<38mg/dL), almost half experienced progression of their plaque burden. In addition, they found that it was not uncommon for patients with LDL-C levels of over 70mg/dL to experience plaque regression. This suggests that merely relying on an LDL-C level to predict plaque burden is insufficient.

Cortical Neuronal Damage in Atherosclerotic Large Artery Disease

Russell Mitesh Cerejo, MD

Yamauchi H, Kagawa S, Kishibe Y, Takahashi M, Higashi T. Progressive Cortical Neuronal Damage and Chronic Hemodynamic Impairment in Atherosclerotic Major Cerebral Artery Disease. Stroke. 2016

In their paper, the authors set out to determine whether selective cortical neuronal damage manifests as a decrease in BZRs in the normal appearing cerebral cortex of patients with atherosclerotic ICA or MCA occlusive disease, and furthermore whether these changes can be correlated with chronic hemodynamic impairment at baseline or hemodynamic deterioration. They studied 80 patients with atherosclerotic ICA or MCA disease with 17 having TIAs, and 38 having completed stroke.

The authors found that the BZR index in 40 patients was increased during follow-up (mean 26±20 months). In multivariable logistic regression analyses, increases in the BZR index were associated with the decreased cerebral blood flow at baseline and an increased oxygen extraction fraction during follow-up. They hypothesize that misery perfusion at baseline is associated with subsequent development of ischemic cortical neuronal damage. The contribution of the increased BZR index at baseline suggests that patients with misery perfusion have already suffered some ischemic cortical neuronal damage and may be at particular risk for progressive cortical neuronal damage.
They also found that increases in the oxygen extraction fraction during follow-up were associated with a lack of statin use. They suggest that revascularization procedures can improve hemodynamic impairment and thus may be beneficial to patients vulnerable to selective neuronal damage. 


Asian patients with paroxysmal AFib have lower stroke risk than those with sustained AFib

Ilana Spokoyny, MD

Takabayashi K, Hamatani Y, Yamashita Y, Takagi D, Unoki T, Ishii M, et al. Incidence of Stroke or Systemic Embolism in Paroxysmal Versus Sustained Atrial Fibrillation: The Fushimi Atrial Fibrillation Registry. Stroke. 2015 

Afib is a major risk factor for stroke, and oral anticoagulation (OAC) has been shown to be the most effective preventative treatment. Since OAC is associated with increased bleeding risk, efforts have been made to stratify risk by identifying patient characteristics which increase the risk of stroke in AFib patients (CHF, Hypertension, Age, Diabetes, etc.). Until recently, paroxysmal and persistent AFib were reported to have the same associated risk of thromboembolic events, and the guidelines recommend OAC treatment for prevention, regardless of the type of AFib. It does seem intuitive that someone in constant AFib for years would have a higher risk of stroke than a patient who had half an hour of AFib after surgery. Several recent studies have shown a lower stroke rate in paroxysmal AFib compared to sustained AFib. There is also limited data in Asian patients, so some equipoise existed which prompted this study. 

Data was obtained from the Fushimi AF Registry, whose only inclusion criteria was the documentation of AF on a 12-lead EKG or Holter monitoring at any time. Paroxysmal AF was defined as lasting under 7 days; persistent AF lasted longer than 7 days was but able to be terminated by medication or electrical cardioversion; permanent AF was refractory to medical or electrical cardioversion. Persistent and permanent AF were combined in this study as SAF (sustained AF), due to difficulty differentiating these in daily clinical practice. Multivariate analyses were performed for the vascular risk factors in both the CHADS2 and CHA2DS2-VASc scores, and also included the variable of “low body weight” (which was shown to correlate with stroke risk in Japanese AF patients).

3304 patients were included, split about evenly between paroxysmal and sustained AF. PAF patients were younger, more often symptomatic (palpitations), and less likely to have a history of stroke, CHF, or CKD. The CHADS2/CHA2DS2-VASc scores were lower in PAF patients, and there was a lower rate of OAC use among PAF patients at any age or CHADS2 score compared to SAF patients. About 10 percent of PAF patients progressed to SAF, and there was a significantly higher rate of stroke/systemic embolism among those who progressed compared to those who did not. PAF patients had lower risk of stroke/systemic embolism overall, and this persisted when the patients were stratified by OAC use and CHADS2 score. On multivariate analysis, PAF was an independent risk factor for a reduced risk of stroke/systemic embolism, while age ≥75 and history of stroke portended increase risk. PAF remained significant even when CHADS2 and CHA2DS2-VASc criteria were added to the model, as well as when age was added as a continuous variable. Low body weight (≤50kg) was indeed an independent risk factor for increased risk of stroke/systemic embolism.

This study did not measure the AFib burden, which may be the key to risk stratification in AFib. The other Asian registry mentioned, J-RHYTHM, did not enroll patients who maintained sinus rhythm for over a year, while this registry did. J-RHYTHM showed more thromboembolism in SAF patients, but this difference disappeared after adjusting for vascular risk factors. Neither registry used long-term monitoring or included AF burden as a variable. In this study, patients with PAF were more likely to receive anti-arrhythmic drugs or catheter ablation, which may have decreased AF burden and stroke risk.

In order for these results to influence the guidelines on treatment of PAF, it would have been important to compare the anticoagulated PAF to non-anticoagulated PAF, as well as to age and risk-matched non-AFib patients. However, since this was a registry, it is difficult to make a comparison between treated and untreated patients as there are other contributing factors which influence the initiation of OAC. This can be seen here, as the incidence rate of stroke/systemic embolism per 100 person-years in OAC-treated patients is higher than in untreated patients (and this difference persists for ischemic stroke).

The discrepancies between PAF and SAF shown in this observational registry study are important and may lend evidence to the idea that AF burden is associated with stroke risk. As monitoring technology advances, we are identifying PAF earlier and more often (and committing patients to OAC sooner), but we also have the capability to quantify the AFib burden. Despite guidelines recommending OAC for PAF, only 39% of patients in this registry were prescribed OAC. So, it would be ethical to perform a study of PAF patients which quantifies AFib burden (via wearable or implanted monitor) and randomizes those with low vascular risk scores and a low AF burden to antiplatelet therapy. This would be very useful in creating safe and evidence-based recommendations on the need for OAC therapy in PAF patients.

Ischemia in Intracerebral Hemorrhage: The Blood Pressure Conundrum

Michelle Christina Johansen, MD

Gioia LC, Kate M, Choi V, Sivakumar L, Jeerakathil T, Kosior J, et al. Ischemia in Intracerebral Hemorrhage Is Associated With Leukoaraiosis and Hematoma Volume, Not Blood Pressure Reduction. Stroke. 2015

What corresponds to DWI lesions in intracerebral hemorrhage (ICH)? Is this reflective of true ischemic injury? The INTERACT trials refocused the neurointensivist on aggressive blood pressure control as this was shown to correlate with decreased hematoma growth and better functional outcomes. But is there a downside to controlling the blood pressure too aggressively? Does this result in ischemic insult or DWI changes on MRI? 




Dr. Gioia et al., sought to identify the frequency of ischemic lesions in primary ICH patients with the hypothesis that larger hematoma volumes and blood pressure reduction are associated with the presence DWI lesions. 

The Canadian based study retrospectively identified 117 ICH patients who underwent DWI within 14 days of symptom onset. Hematoma/perihematoma edema volumes were measured using planimetric techniques. Perihematoma and remote DWI lesion volumes were measured using apparent diffusion coefficient. The investigators established two thresholds to define the amount of ischemic damage; moderate (<730×10-6mm/s) and severe (<550×10-6mm/s). Acute blood pressure changes over the first 24 hours were calculated using the formula admission systolic blood pressure (SBP) minus nadir SBP. Mean age of the population was 65 and 52% were male. Hypertension was the most common cause indicated in the medical record for ICH (45.7%).

The authors, after controlling for age, baseline NIHSS and perihematoma edema volumes, found that perihematoma DWI lesions were independently associated with larger hematoma volumes. There was no association between changes in blood pressure (maximal systolic blood pressure drop) and the presence of DWI lesions. They did find remote DWI lesions in 17 patients. For the remote lesions, once again no relationship was found with blood pressure, but they did note these patients were more likely to use antiplatelet drugs, have a history of ICH and larger leukoaraiosis (hyperintensity on FLAIR) volumes.

How should the results of this study be applied? The mean time to MRI was 2 days which is a consideration when drawing conclusions between volumes and the presence of DWI lesions. That being said, there was excellent interreliability among the radiologists and they accounted for edema volume which should help mitigate over calculation of hematomal volumes. Blood pressure data was missing in 4 patients with perihematoma DWI lesions in a relatively small cohort of patients (38). Would a larger sample size have led to different results? How does the fact that the majority of patients had lobar ICH affect the data? Thus we are still unable to define the relationship between blood pressure control and patients with cerebral amyloid angiopathy.

Where does the vascular neurologist go from here when faced with perihematoma diffusion restriction after ICH? In ischemic injury, we allow for permissive hypertension to provide critical blood flow to the penumbra yet research supports aggressive blood pressure control for ICH. The authors of this study did not find a relationship between aggressive blood pressure control and ischemia but it cannot be ruled out. In their discussion, they suggest that larger hematomas be treated more conservatively with respect to blood pressure control but appropriately caveat this statement with the need for larger clinical trials. While the answer remains unclear, we must strive to clarify this issue so that our patients can receive the best medical care based on evidence.

Poststroke Angiogenesis: Architect and/or Demolition Crew?

Mark N. Rubin, MD

Greenberg DA. Poststroke Angiogenesis, Pro: Making the Desert Bloom. Stroke. 2015


Adamczak J, and Hoehn M. Poststroke angiogenesis, Con: The Dark Side of Angiogenesis. Stroke. 2015


This installment of the “Controversy” series involves what to make of post-stroke angiogenesis: the hypoxia-triggered generation of new capillaries after a stroke of any subtype. The fact that angiogenesis exists in the post-stroke setting–experimental and clinical–is not the point of debate but to what degree this process influences patient outcome. Experimental, pathological case study and treatment trial data exist in this field, but a fundamental clinical question remains unanswered: does manipulating this pathophysiologic process make patients better?



Dr. Greenberg from the Buck Institute for Research on Aging proposes that post-stroke angiogenesis is a viable therapeutic target, mostly because it is fairly well understood at a biochemical level, broadly applicable across the patient population and there are myriad promising biochemical targets in the process that have not yet been investigated. Furthermore, there is the optimistic view that angiogenesis allows for more rapid clearing of ischemic debris, setting a clean slate for post-stroke neuronal reorganization (e.g., functional recovery). Prof. Doctors Adamczak and Hoehn from the Max Planck Institute argue that angiogenesis is more demolition crew than architect. While not arguing that restoration of cerebral blood flow is beneficial to neuronal tissue, they point out the double-edged sword of pro-angiogenic factors (namely Vascular Endothelial Growth Factor, better known as VEGF), which also promote increased cerebral edema which is injurious to brain. They cite evidence that supporting anti-angiogenic factors actually decreases infarct volume.

Dr. Liu from UCSF ties the debate together with a resounding “you’re probably both right but we don’t know enough in general. Plus, how does this all relate to collateralization, which is so hot right now?” She also suggests a careful marriage of nanotechnology and pharmacotherapy may help deliver the right mix of biochemicals–whichever those may be–to the right place at the right time, thus mitigating the known inefficacy and/or risks of systemic delivery of pro-angiogenic factors.

Read on for details of the Basic Science Controversy In Stroke!