External Validation of Six Scores Differentiating Atherosclerotic vs. Embolic Large Vessel Occlusion

In an external validation study of 91 patients, the REMIT score demonstrated the most robust and statistically significant ability to differentiate intracranial atherosclerotic disease-related large vessel occlusion from embolic occlusion, supporting the superiority of scores incorporating imaging features over those relying solely on clinical variables.

The Gist

The Big Picture: The "Traffic Jam" Mystery

Imagine your brain's blood vessels are a complex highway system. Sometimes, a massive traffic jam (a stroke) happens because a giant truck (a blood clot) rolled off a truck on a different highway and got stuck in the middle of the road. This is called an Embolic Stroke.

Other times, the road itself is crumbling. The asphalt is old, cracked, and narrowing over years. Eventually, the road gets so narrow that traffic stops right there. This is called Atherosclerotic Stroke (caused by plaque buildup).

Why does this distinction matter?
If the problem is a truck (embolism), you just need to tow the truck away (remove the clot), and the road is clear.
If the problem is crumbling asphalt (atherosclerosis), removing the debris isn't enough. The road is still broken and narrow. If you just clear the jam, the road might collapse again immediately. You might need to build a bridge (stent) or fix the road surface (angioplasty).

The Problem: When a patient arrives at the hospital in a panic, doctors have to decide instantly which type of "traffic jam" they are dealing with. They need a quick way to tell the difference between a "rogue truck" and "crumbling asphalt."

The Study: Testing the "Detective Tools"

Over the years, doctors have created six different "checklist scores" (like a detective's notebook) to help them guess the cause of the stroke before they even start the surgery. These checklists look at things like:

• The patient's age and medical history (Do they have heart issues?).
• Blood test results.
• Pictures of the brain (CT scans) to see what the blockage looks like.

The Goal of This Paper:
The researchers wanted to see if these six checklists actually work when used by a different group of doctors in a different hospital. They took data from 91 real patients who had just undergone stroke surgery and ran the numbers through all six checklists to see which one was the best detective.

The Results: Who Won the Detective Contest?

Think of the six checklists as six different detectives trying to solve the case. The researchers measured how good they were using a score called the "AUC" (think of it as a grade out of 100, where 100 is perfect).

1. The Winner (REMIT): This checklist got a grade of 79/100. It was the most reliable at telling the difference between the two types of strokes. It paid close attention to things like heart failure markers and the specific shape of the blockage.
2. The Runner-Up (Score-ICAD): This one got a 71/100. It was also quite good.
3. The "Almost" (ISAT): This one got a high 87/100, but because there were very few cases of the specific type of stroke it was designed for (back-of-the-brain strokes), the result wasn't statistically "proven" yet. It's like a detective who got the right answer but only had one case to solve.
4. The Strugglers (ABC2D, ATHE, ICAS-LVO): These three got grades between 46 and 63. They were basically guessing. They couldn't reliably tell the difference between a "rogue truck" and "crumbling asphalt."

The Big Discovery: Pictures Beat Stories

The most interesting finding of the study wasn't just which checklist won, but why it won.

• The "Story" (Clinical History): Things like "Does the patient have diabetes?" or "Do they have high blood pressure?" were weak clues. They didn't help much in telling the difference.
• The "Photo" (Imaging): The clues found in the brain scans were superheroes.

The study found that specific visual patterns on the CT scan were the strongest predictors. For example:

• The "Tapered" Sign: If the blockage looks like it slowly narrows down to a point (like a carrot tip), it's likely crumbling asphalt (atherosclerosis).
• The "Non-Culprit" Stenosis: If the doctor sees other roads in the brain that are also narrowed, it suggests the whole highway system is aging and crumbling, not just one spot.

The Analogy:
Imagine trying to guess if a car broke down because of a flat tire (embolism) or because the engine is old and worn out (atherosclerosis).

• Clinical History is like asking the driver, "Do you usually drive fast?" (Not very helpful).
• Imaging is like popping the hood and looking at the engine. If you see oil everywhere and rusted parts, you know it's an old engine problem, regardless of how fast the driver usually goes.

The Conclusion: What Should We Do?

The researchers concluded that while these checklists are helpful, the best way to know what's going on is to look closely at the pictures of the blood vessels, not just the patient's medical history.

They suggest that in the future, doctors should rely more on the "visual clues" (like the shape of the blockage and other narrowed arteries) to decide how to treat the stroke. This helps ensure that if the road is crumbling, they fix the road, not just clear the traffic.

In short: To solve the mystery of a brain stroke, don't just listen to the story; look at the crime scene photos. The pictures tell the truest story.

Technical Summary

1. Problem Statement

Acute ischemic stroke caused by Large Vessel Occlusion (LVO) is increasingly treated with mechanical thrombectomy (MT). However, the underlying etiology is heterogeneous, primarily falling into two categories:

• Embolic LVO (EMB-LVO): Caused by a clot traveling from a distant source (e.g., heart).
• Intracranial Atherosclerotic Disease-related LVO (ICAD-LVO): Caused by in-situ thrombosis on a pre-existing stenotic plaque.

Clinical Challenge: Differentiating between these two etiologies in the acute setting is critical because their management differs significantly. ICAD-LVO carries a high risk of reocclusion and often requires rescue therapies (angioplasty, stenting, or aggressive antiplatelet therapy), whereas EMB-LVO is typically treated with thrombectomy alone.
Gap: While several prediction scores (scales) have been developed to distinguish ICAD-LVO from EMB-LVO using clinical and imaging variables, their external validity (performance in independent, real-world cohorts) remains unproven. Furthermore, the relative contribution of individual components within these scores has not been systematically analyzed.

2. Methodology

• Study Design: Retrospective analysis of prospectively collected data from a multicenter stroke registry.
• Setting: Two centers in Japan (Jikei University Hospital and Jikei University Kashiwa Hospital).
• Timeframe: June 2021 to March 2025.
• Participants:
  • Inclusion: Adults with acute ischemic stroke due to LVO who underwent mechanical thrombectomy and had complete clinical/imaging data for six specific scores.
  • Exclusion: Patients without contrast-enhanced CT, those with tandem lesions, or those with only pre-procedural MRI.
  • Final Cohort: 91 patients (18 ICAD-LVO, 73 EMB-LVO).
• Gold Standard Definition:
  • ICAD-LVO: Defined as >50% residual stenosis at the culprit site after thrombectomy.
  • EMB-LVO: All other cases.
• Scores Validated: Six prediction scales were evaluated:
  1. ISAT: For posterior circulation only.
  2. REMIT: General LVO (embolism focus).
  3. ABC2D: General LVO.
  4. ATHE: Anterior circulation only.
  5. ICAS-LVO: Requires intra-procedural angiography.
  6. Score-ICAD: Anterior circulation only.
• Statistical Analysis:
  • Discriminative performance was assessed using the Area Under the Receiver Operating Characteristic Curve (AUC).
  • Individual components of the scores were analyzed for their independent predictive value using univariate odds ratios (OR).

3. Key Contributions

• First External Validation: This study provides the first external validation of these six specific prediction scores in a consecutive, real-world cohort, moving beyond the derivation datasets.
• Component Dissection: The authors deconstructed the scores to identify which specific variables (clinical vs. imaging) actually drive the prediction of ICAD-LVO, rather than just relying on the composite score.
• Imaging vs. Clinical Emphasis: The study systematically demonstrates that imaging features are superior to clinical history in predicting ICAD-LVO etiology.

4. Results

A. Demographics and Baseline:

• ICAD-LVO prevalence was 20% (18/91).
• Significant Differences: The ICAD-LVO group had significantly lower rates of atrial fibrillation (11% vs. 41%, p=0.017) and lower BNP levels (55.1 vs. 121.4 pg/mL, p=0.037) compared to the EMB-LVO group.

B. Performance of Prediction Scores (AUC):

• REMIT Scale: Demonstrated the most robust and statistically significant discrimination (AUC 0.793, 95% CI 0.676–0.911, p < 0.001).
• ISAT Score: Showed the highest numerical AUC (0.870) but did not reach statistical significance (p = 0.064), likely due to the small sample size of posterior circulation cases (n=18).
• Score-ICAD: Significant predictor (AUC 0.707, p = 0.013).
• Poor Performers:
  • ABC2D (AUC 0.627, p = 0.095).
  • ATHE (AUC 0.600, p = 0.230).
  • ICAS-LVO (AUC 0.465, p = 0.650) – performed worse than random chance.

C. Analysis of Individual Components:

• Clinical Variables: Only the absence of atrial fibrillation was significantly associated with ICAD-LVO (OR 6.600). Other factors like NIHSS, blood pressure, and glucose were not predictive.
• Imaging Variables: Imaging findings were the strongest predictors. The most significant factors included:
  • Non-culprit stenosis: Highest OR (21.071, p < 0.001).
  • Multiple arterial stenosis: OR 20.571 (p = 0.009).
  • Border zone infarction: OR 10.143 (p = 0.011).
  • Tapered sign: OR 4.295 (p = 0.031).

5. Significance and Conclusion

• Validation of Imaging-Centric Models: The study confirms that prediction models incorporating specific imaging features (particularly those reflecting diffuse atherosclerotic burden like non-culprit stenosis and border zone infarction) outperform those relying heavily on clinical variables.
• Pathophysiological Insight: The results support the concept that ICAD-LVO is a distinct pathophysiological entity characterized by chronic hemodynamic compromise and diffuse atherosclerosis, rather than an isolated embolic event.
• Clinical Implication: Accurate mechanism inference requires comprehensive imaging assessment of the entire intracranial arterial tree, not just the occlusion site. The REMIT and Score-ICAD scales appear most reliable for external use.
• Future Directions: The authors suggest that future models should integrate perfusion imaging (e.g., mismatch ratios, Tmax) to further refine the identification of ICAD-LVO, as these metrics reflect vascular reserve and collateral circulation associated with chronic stenosis.

Limitations: The study was retrospective, had a modest sample size for ICAD-LVO (especially in posterior circulation), and relied on angiographic rather than pathological confirmation of etiology.