Complications following endovascular clot retrieval for stroke

Endovascular clot retrieval is relevant to "Acute cerebrovascular injury" from Section 2.1.8 in the second edition of the CICM Syllabus for the Second Part Examination. More specifically, the complications of ECR are relevant to answering Question 1 from the second paper of 2025, which asked for them (at least six of them),  as well as an approach to the assessment of a disappointingly unconscious patient after ECR. The material that follows has been separated from the principal chapter on endovascular clot retrieval on the grounds that it bears for the examination candidate a certain practical usefulness, which the parent chapter, for the most part, does not.

In summary:

In terms of published evidence, nothing seems to bring everything together under one title, and the trainees must resort to scanning disparate sources, eg. Shapiro et al (2020) for access complications, Wang et al (2017) for seizures, Li et al (2020) for reocclusion, and so on. The Australian paper on the ICU management of the post-ECR patient by Ma Moynihan and Donaldson (2019) remains a solid recommendation, if you only have time to read one paper.

Post-procedure complications in ECR

The most common must surely be some variation on the theme of failed procedure or acces site injury. 

The intensivist is faced with a patient who continues to suffer all the previously noted deficits of their stroke, and is likely on the way to develop some new ones. In addition to this, they have recently had some large artery punctures in the context of recent fibrinolysis, and their cerebral vessels have been interfered with  mechanically by probes wires and contrast media. In short, it should be easy to rattle off a list of complications in response to the question "list the complications". The reviews by Balami et al (2018) and Krishnan et al (2021) are particularly good for this:

• Things you can detect early, and do something about:
  • Access site vascular injury: 
    • Pseudoaneurysm or haematoma at the access site
    • Dissection at the access site
    • Retroperitonal haematoma
  • Malignant MCA infarction
  • New embolisation
• Things which will just happen, and which have no satisfying solution: 
  • Contrast reaction or nephrotoxicity
  • Intracranial haemorrhage
  • Dissection distal to the site
  • Cerebral oedema

What "failed" means, of course, is a matter for the pedants (was reperfusion impossible in spite of multiple attempts, or did we give up because groin access was too hard?). Moreover the authors of papers typically do not classify an unsuccessful procedure as a complication, and when they do, they cannot agree on how to define it (eg. in terms of TICI, how much flow is enough?) TICI, or more recently mTICI and eTICI, is a system to classify cerebral reperfusion as:

• Grade 0: No perfusion
• Grade 1: Little or slow distal reperfusion
• Grade 2a: Reperfusion of less than half of the occluded target artery
• Grade 2b: More than half of the occluded target artery
• Grade 3: Complete antegrade reperfusion

This was all laid down in a consensus statement by Zaidat et al (2013), following the original paper by Higashida & Furlan (2003) who first called for an end to the application of the 1985 TIMI scale to cerebral vessels, on the grounds that the directly visualised passage of contrast in cerebral vessels is somehow fundamentally different to the flow in the coronaries. The main argument for having a completely separate scale was that the final infarct volume has much more bearing on the functional outcome in the former, i.e. a damaged motor cortex is much harder to live with than a regional wall motion abnormality.

Different ways in which ECR can fail to achieve reperfusion

Leslie-Mazwi et al (2017) list the following failure options:

• It remains occluded. This is not a rare problem: 90% success rate means you are unsuccessful in 10%. Specifically, Kaesmacher et al (2018) found no reperfusion in 10.6% (TICI 0 or 1, i.e. either zero flow or  sluggish contrast penetration into the distal branches). It does not help that nobody can agree how much flow is enough - Heider et al (2020) laments that "a reperfusion result of mTICI 2a is sometimes regarded as success, sometimes as failure", which changes the statistics significantly as fully a third of their failed cases were in this category (including vs. excluding them changed the failure rate from 16.8% to 11.3%). From an analysis of HERMES trial data by Liebeskind et al (2019), we can see that reperfusion success is a fairly flat bell curve:

  Final eTICI	Reperfusion description	Proportion
  0	No reperfusion	8%
  1	No distal branch filling	3%
  2a	Less than half the territory reperfused	14%
  2b50	50-66% reperfusion	14%
  2b67	67-89% reperfusion	30%
  2c	90-99% reperfusion	23%
  3	Complete reperfusion	9%

   What is the point of this digression? For the context of the patient failing to awaken following ECR, a "successful" reperfusion may be called on a eTICI 2b50, which leaves about 50% of the patients with a modified Rankin score of 3-6 (moderate to severe disability).

• It reoccluded. 3-9% will go in this direction within the first 24 hrs, according to Li et al (2020). A longer onset-to-reperfusion time seems to be the main risk factor. The stroke being cardioembolic seems to be protective, as does the use of antiplatelet agents prior to the stroke. Large occluded vessels seem to also be a favourable feature (specifically M1).
• It was either already dissected, or you just dissected it. Hard to know; dissection can be a valid explanation for what appeared to be an embolic occlusion, or it could be a complication inflicted during the attempt. The latter is usually detected and reacted upon, typically with some kind of stent device.
• Neither is it occluded, nor is the brain viable. The ECR was heroic but late, and the brain remains neurologically ungrateful. One might ask, how can a procedure that is 90% successful be observed to produce poor functional outcomes in about 50% of patients? But that's exactly the finding of  Shahid et al (2022). The factors associated with this did not seem especially modifiable from this study, consisting of things like sex, age, background history, and the technical challenges of the procedure.

Complications of successful reperfusion

Despite the apparent success of reperfusion and clinical improvenmentof the injured brain, problems can develop which can frustrate or reverse that recovery:

• Haemorrhagic transformation.  Hao et al (2017) looked through all the major publications on the outcomes following endovascular clot retrieval and reported that among the 1,499 reported patient cases,  the risk of intracranial haemorrhage was substantially higher in the endovascular group. That's right, medical management actually carried a much lower risk of intracranial haemorrhage, only 19% - whereas with the endovascular approach the risk was 35%. However, this is only surprising if one assumes that "medical management" is systemic thrombolysis. In actual fact, Hao et al included trials which grouped "best medical therapy" together with intravenous t-PA, which probably dropped the risk of ICH somewhat. Moreover, even with this caveat, the extra intracranial blood in the interventional groups appears to have gone largely unnoticed by the patients and their families: when only "symptomatic"  stroke was included in the analysis, the numbers were much lower, and very similar (5.6% vs 5.2%).  
• Cerebral oedema can develop, with an ensuing rise in the intracranial pressure and adverse effects of midline shift. Malignant MCA-like management, including decompressive craniectomy, is then floated as an idea. One might counter this enthusiasm by pointing out that, if the brain has swollen quite as bad as this, then one might make the argument that the volume of salvageable tissue was not rescued, and the outcome for this patient will be more like the outcome for untreated stroke (i.e. bad no matter what you do). It would be tempting to separate this category into a nest of associated points (mass effect, midline shift, herniation) but many examiners would agree that these are all slices from the same salami and are therefore not worth awarding individual marks. In other words, listing six different versions of cerebral oedema for Question 1a would be awarded only 0.5 marks, assuming it was half a mark per listed cause.
• Seizures  are common (in 5.5% for Jung et al, 2012), which is slightly more common than in all strokes (3.3% for Wang et al, 2017)  but not common enough for routine prophylaxis (Prabhakaran et al, 2026). One must of course point out that only a reperfused brain would be having seizures, so we could even view this as some kind of positive sign.
• Vascular access site complications are seen in a substantial proportion of these patients as they are often full of antiplatelet agents at the time of interventional access sheath removal. The total rate of serious problems found by Shapiro et al (2020) was 1.67%, which they described as "not low", and which mainly consisted of groin haematoma and pseudoaneurysm. The authors pointed out that this is not unique to ECR and seems to be similar to the rates of transfemoral access complications in interventional cardiology.

Assessment of the patient who remains unconscious following endovascular clot retrieval

Question 1 from the second paper of 2025 presented the candidates with a patient who, after six hours off sedation following ECR, remains unresponsive on the ventilator. Apart from the list of possible procedure complications listed above, one could include:

• Prolonged anaesthetic drug effect, eg. where organ function hinders elimination
• Concomitant metabolic reasons,  eg. hyponatremia
• Unrelated intracranial pathology,  eg. SDH due to a stroke-related fall, or venous sinus thrombosis

History

As is generally the case, the history aspect of the assessment is mostly a catalogue of risk factors or protective features that raise or lower your suspicion of specific complications. And to save time, you can continue to pontificate on the complex interplay of pre-test probabilities raised by the history while you escort the patient down for the head CT which they will inevitably get irrespective of the history or risk profile.

This information could include: 

• Duration from stroke onset to reperfusion: this is a risk factor for multiple complications, notably reocclusion
• Initial NIHSS score and GCS at presentation
• Aphasia pre-procedure (i.e. they are not obeying commands because they do not understand them)
• Volume of the stroke on pre-procedure imaging (predictive of malignant MCA syndrome)
• History of sedation, anticoagulation, antiplatelet drugs, prehospital thrombolysis
• Procedural complexity, i.e. how many passes did they take (each time exposing the cerebral arteries to the risk of perforation)

Examination

• Surely some sort of neurological examination would be in order. At its most basic, this should focus on:
  • Reassessment of the GCS
  • Cranial nerve exam looking for features of herniation
  • Motor examination looking for worsening of the pre-stroke power and tone
  • Features of nonconvulsive seizures
• This examination should be serial

Investigations

• Imaging, obviously. A noncontrast CT would immediately exclude a lot of the important differentials, a perfusion CT would eliminate reocclusion and dissection, and whatever remains requires an MRI which is not usually available instantaneously but which can be useful in determining residual infarct size and any new embolic phenomena.
• Bloods contribute little, but exclude differentials which are far down the list of possibilities, like a new coagulopathy, hypoglycaemia, electrolyte derangement, et cetera.

Imaging weirdness following endovascular clot retrieval

Gurbani et al (2025) walks the reader through the colourful world of brain perfusion scanning to educate them about the sneaky errors of interpretation and the unique imaging findings that can develop following reperfusion. This crosses into radiology and probably requires a chapter all to itself, which means it is deprioritised in the author's to-do list (given that perfusion scans are unlikely to be inflicted on Second Part exam candidates).