Frequency, Determinants, and Outcomes of Emboli to Distal and New Territories Related to Mechanical Thrombectomy for Acute Ischemic Stroke

Endovascular thrombectomy (EVT) is now a well-established therapy of proven benefit for acute ischemic stroke (AIS) due to large vessel occlusion.¹ However, despite high rates of recanalization of target occlusions with EVT, excellent, nondisabled outcomes are achieved in only a minority of patients.² Clot fragmentation and distal embolization during extraction is a known complication of EVT that may unfavorably influence clinical outcome. However, the frequency, determinants, and outcomes of distal embolization during EVT is poorly characterized.

Distal embolization of fragmented target thrombi was categorized into 2 broad classes in a recent international consensus statement: (1) emboli to new territory (ENT) and (2) emboli to distal territory (EDT).³ With ENT, emboli typically escape after having been pulled below the initial target occlusion vessel and enter a new arterial territory not previously exposed to ischemia. A common example is new anterior cerebral artery embolic occlusion after EVT for a target middle cerebral artery occlusion. With EDT, emboli escape typically during device engagement with the clot in the initial target vessel and travel to more distant branch vessels. EDTs cause persisting or intensified ischemia in brain regions already exposed to ischemia from the initial occlusion.

To date, studies of embolization during EVT have predominantly used digital subtraction angiography (DSA) and magnetic resonance (MR) diffusion-weighted imaging (DWI) to detect distal emboli, but these techniques have important limitations. For ENT, catheter angiography may not always reliably distinguish when a vessel cutoff reflects distal embolus versus competitive hemodynamic flow patterns or vasospasm, and DWI only identifies emboli that produce cerebral infarction, rather than the full spectrum of emboli that may also cause subinfarctive ischemia. For EDT, DSA and DWI are even less informative. DSA generally cannot distinguish whether a distal vessel occlusion (DVO) in the target territory is a preexisting thrombus that was present at that site before the procedure versus a new embolic thrombus to that site that occurred during the procedure. Also, strict immobilization is needed to recognize distal emboli that are more difficult to discern than proximal emboli. Similarly, DWI cannot distinguish when ischemic injury occurred in a distal branch territory as a result of ischemia caused by the initial target occlusion or a later EVT-related embolus.

Susceptibility-weighted MR imaging (MRI) techniques have several advantages as a method to visualize distal emboli after EVT, especially EDT not well characterized by other techniques. Gradient-recalled echo (GRE) MR sequences directly visual distal emboli, evident as blooming artifacts within distal vessels, permitting differentiation from vasospasm and hemodynamic flow cutoffs and recognition of DVOs not leading to infarction. In addition, comparison of pre- and postprocedural images enables distinction of distal emboli present before the procedure from emboli occurring during the procedure. We, therefore, used serial GRE-MRI to investigate the frequency, determinants, and outcomes of patients with EVT-related emboli.

Among the 50 patients meeting study entry criteria, mean age was 70 (±16) years, 44% were women, and median National Institutes of Health Stroke Scale score 15 (interquartile range [IQR], 8–19). Time from last known well to arterial puncture was median 6 hours 41 minutes (IQR, 4 hours 4 minutes to 10 hours 38 minutes). All patients were treated with stent retrievers, aspiration thrombectomy, or both, including stent retrievers alone in 70%, aspiration alone in 8%, and stent retriever and aspiration in 22%. In 20% of patients, additional endovascular recanalization techniques were also used, including intracranial angioplasty without stenting in 2%, cervical angioplasty/stenting in 14%, and wire maceration in 4%. Substantial reperfusion (modified Thrombolysis in Cerebral Infarction score 2b-3) was achieved in 84%.

Pretreatment MRIs were performed a median 58.5 minutes (IQR, 45–73) before arterial puncture. Post-treatment MRIs were obtained at 2 time points in 84% of patients and 1 time point in 16%. Puncture-to-first-post-treatment-MRI time was median 6 hours 2 minutes (IQR, 4 hours 51 minutes to 8 hours 32 minutes). Puncture-to-second-post-treatment-MRI time was median 25 hours 41 minutes (IQR, 23 hours 45 minutes to 27 hours 14 minutes).

New emboli evident on postprocedure GRE were present in 22% (11 of 50) patients. All of the GRE-evident emboli were in the field of the initial target occlusion (EDT) and none in a new arterial territory (ENT; example in Figure 1). Among patients with EDT-SVS, the emboli were single in 91% and multiple in 9%. Among the 12 total EDT-SVS, 33% were associated with new diffusion abnormalities in fields supplied by the compromised distal artery, 25% were in distal arteries supplying tissues that already evidenced ischemic diffusion abnormalities before the thrombectomy procedure, and 42% were not associated with diffusion abnormalities. Patients with GRE-MRI EDT-SVS nominally but nonsignificantly more often had DVOs evident on end-of-procedure catheter angiography, 82% (9 of 11) versus 56% (22 of 39; P=0.13).

Diffusion abnormalities in new territories were found in 11 (22%) patients, including 7 patients with 1 INT-DWI, 2 with 2, 1 with 4, and 1 with 5 (example in Figure 2). Detailed anatomic locations of the INT-DWIs are shown in Table I in the Data Supplement. Among the 7 anterior circulation patients with INT-DWI, INT were ipsilateral to the target occlusion in 3, contralateral in 3, and bilateral in 1. Among 4 posterior circulation patients with INT-DWI, INTs were located in vascular territories proximal to the target occlusion in 3 and in the anterior circulation in 1.

In this study of consecutive large vessel occlusion AIS patients undergoing EVT, MRI-delineated new embolic events associated with the EVT were common, occurring in about 2 in every 5 patients. New EDT beyond the target occlusion occurred in about 1 in every 5 patients; and new infarcts in previously uninvolved territories occurred in about 1 in every 5 patients, with only 1 in 50 patients having both types. Patient features consistently associated with new EDT, on both multivariate and classification and regression tree analysis, were intravenous tPA pretreatment (odds increased 5-fold) and lower pretreatment DBP (odds increased 10% per every 1 lower mm Hg). Higher rates of INT were associated with vertebrobasilar target occlusion location (odds increased 7-fold), while lower rates were associated with SVS on preinterventional imaging (odds reduced 5-fold) and internal carotid artery target occlusion location (odds reduced 8-fold). INTs were associated with reduced hemorrhagic transformation; EDTs were not. Neither INT nor EDT was associated with outcomes of hyperacute injury marker sign, infarct extent at 16 to 48 hours, or disability level at discharge or 3 months.

The rate of distal embolization in this series is comparable to a recent series of susceptibility-weighted imaging–based detection of DE.¹⁰ The association of EDT-SVS with bridging thrombolysis is unsurprising, as prior studies have reported increased rates of periprocedural thrombus fragmentation after lytics, as detected on DSA.^11,12 The association of ICA target occlusion with EDT-SVS suggests retrieval of thrombi from the proximal intracranial vasculature may require additional caution, as the larger target thrombi pose greater embolic risk. The inverse association of ICA target occlusion with INT-DWI likely reflects that, when control over a thrombus is lost during retrieval from the ICA target, the likely result is embolization into the original territory, as there rarely are branch points into other territories proximal to the ICA target. The association of vertebrobasilar target occlusions with INT-DWI, in part, likely reflects larger target thrombi with greater fragmentation risk. Treatment technique aspects likely also contribute to the higher embolism frequency with vertebrobasilar target occlusions, as balloon guide catheters are used less and aspiration alone more commonly than in anterior circulation occlusions. The predictive value of DBP <66 mm Hg in multivariate models highlights the importance of hemodynamic factors in the clearance of distal emboli, so called washout.¹³ DBP decreases reduce intravascular pressure that clears residual clots from the arterial tree. In addition, reduced blood pressure lowers collateral flow, which may decrease effective circulation of lytic plasma-based factors to downstream emboli.¹⁴ The association of atrial fibrillation and target occlusion SVS with EDT-SVS suggests a possible role for clot composition in the likelihood of clot fragmentation during procedural manipulation. Clot composition has been demonstrated to vary by stroke etiology¹⁵ and SVS-positive thrombi shown to be erythrocyte rich.¹⁶ Overall, the relationships demonstrated here suggest that several parameters, including clot characteristics, hemodynamics, collaterals, clot size, and location, may affect periprocedural embolization.

The rate of INT in the current study was generally higher than in prior series and is likely to be more accurate, as the current study uniquely employed more sensitive DWI in all patients for follow-up imaging.^17–19 The association of INT-DWI with reduced hemorrhagic transformation was unexpected but likely has several sources. In part, it reflects the association of INT-DWI with posterior circulation target occlusions, which less often have post-thrombolytic hemorrhagic transformation, and the inverse association of INT-DWI with ICA target occlusions, which more often have post-thrombolytic hemorrhagic transformation.^20–23 Additionally, INT-DWI patients had a nominal decrease in the rate of substantial reperfusion, which may have played a protective role against hemorrhagic transformation.²⁴ In contrast to prior studies, INT in the current study was not associated with changes in disability level at 3 months.^17,19 Likely the DWI surveillance in the current study detected smaller infarcts than prior investigations, which are associated with less impact upon functional outcome.

It is noteworthy that EDT-SVS was observed less often than DVO-DSA at procedure end, 22% versus 62%. This lower frequency likely has 3 sources. First, some of the DVOs at procedure end seen on DSA were unretrieved portions of the initial target occlusion, rather than new EDT; in contrast, GRE-MR imaging enabled disambiguation of residual original thrombus from new distal emboli. Second, among those that were distal emboli, a substantial proportion are likely to have lysed and resolved by the time of follow-up MRI imaging 3 to 28 hours after procedure completion. Third, some distal emboli will be fibrin-rich, erythrocyte-poor thrombi that do not generate SVS.

This study has several limitations. First, GRE is known to be a less sensitive method of detecting DE as compared with susceptibility-weighted imaging, which offers phase-sensitive information and 3-dimensional rendering that increase its power to detect small distal emboli, and SVS does not show fibrin-rich emboli. In addition, some initial distal emboli may have spontaneously lysed between the end of procedure and the time of the follow-up GRE-MRI. As result, the 22% rate of new DE represents a conservative value but provides a useful minimum rate at which DEs are occurring. It is also important to note that proper evaluation of any susceptibility-weighted sequence for the presence of distal emboli requires detailed knowledge of cerebrovascular anatomy and careful consideration of radiological differential diagnoses including cerebral microbleeds, normal arterial flow voids, and deoxygenated blood within cerebral veins. In addition, susceptibility artifacts due to hemorrhagic transformation may further complicate the identification of EDT-SVS in close anatomic proximity. Importantly, radiological suspicion of EDT-SVS may be supported with additional MRI findings including fluid-attenuated inversion recovery hyperintense vessels (indicative of slow flow) and perfusion delay (Figure 1). Additionally, at our center, images were typically acquired at a slice thickness of 5 mm, which may miss EDT-SVS due to partial volume averaging. Despite these technical limitations, GRE-based detection of EDT-SVS may offer superior clinical applicability, as susceptibility-weighted imaging is not yet recommended per national guidelines regarding acute stroke imaging and management.^25,26 Our study is also limited by its retrospective nature and relatively small sample size.

Among AIS patients treated with modern EVT technology, embolization distal to the target occlusion and INT each occur in about 1 in every 5 cases. Strong predictors of distal embolization are ICA target occlusion, indicating larger thrombus to be manipulated; pretreatment intravenous fibrinolysis, potentially disrupting thrombus structural integrity; and lower DBP, suggestive of impaired embolic washout.

List of UCLA Thrombectomy Investigators: Rodel Alfonso, RN; Allison Arch, MD; Gilda Avila, BS; Mersedeh Bahr Hosseini, MD; Fionna Chatfield, RN; Arun S. Chhabra, MD; Bruce Dobkin, MD; Gary Duckwiler, MD; Lindsey K. Frischmann, DO; Ileana Grunberg, RN; Judy Guzy, RN; Jason Hinman, MD, PhD; Josephine F. Huang, MD; Doojin Kim, MD; David S. Liebeskind, MD; Konark Malhotra, MD; Michael McManus, MD; May Nour, MD, PhD; Neal Rao, MD; Lucas Restrepo, MD; Jeffrey L. Saver, MD; Latisha K. Sharma, MD; Sidney Starkman, MD; Parampreet Singh, MD; Xiannan (Xander) Tang, MD; Jason W. Tarpley, MD; Anita Tipirneni, MD.