Thrombectomy for medium-vessel stroke: what the trials show

In two large randomized trials published together in the New England Journal of Medicine in April 2025, removing the clot with a catheter did not reduce disability at 90 days for stroke caused by a blocked medium or distal artery, compared with medical treatment alone. One of the two trials also reported higher mortality in the thrombectomy group. For a procedure with a large, well-documented effect when a large artery is blocked, that is a striking result, and it is a useful case for seeing how evidence from one group of patients can fail to transfer to a neighbouring one.

Does thrombectomy help when a medium or distal artery is blocked?

In the two trials described here, it did not. This is an explainer about method and evidence, part of the broader stroke outcomes research this site follows. It describes what the trials found. It does not describe what any patient or clinician should do.

A few terms first

An ischemic stroke happens when a clot blocks an artery that feeds the brain. Endovascular thrombectomy (often shortened to EVT) is the procedure that removes the clot directly: a catheter is threaded through the blood vessels up to the blockage and the clot is pulled out, restoring flow.

Where the blockage sits matters. A large-vessel occlusion blocks one of the main arteries at the base of the brain. A medium or distal vessel occlusion blocks a branch further downstream. The DISTAL trial defined these by named arterial segments: the nondominant or codominant M2 segment, or the M3 or M4 segment, of the middle cerebral artery; the A1 to A3 segments of the anterior cerebral artery; and the P1 to P3 segments of the posterior cerebral artery. Downstream branches supply smaller regions of brain.

Both trials measured recovery with the modified Rankin Scale, a 0 to 6 score of everyday function in which 0 means no symptoms and 6 means death. If the scale is new, there is a fuller explanation of the modified Rankin Scale and why stroke trials rely on it.

Why did researchers expect a benefit?

Both papers open from the same starting point. The ESCAPE-MeVO authors wrote that it was unclear whether the "large effect size" of thrombectomy for large-vessel occlusion applies to medium-vessel occlusion. The DISTAL authors described thrombectomy for large-vessel occlusion as known to be safe and effective, and its effect in medium or distal vessels as unclear.

The reasoning that a benefit might carry over is intuitive. The mechanism looks the same: a clot starves brain tissue, and removing it restores blood flow before the starved tissue dies. The race to save that at-risk tissue, sometimes called the ischemic penumbra, is the logic behind every reperfusion treatment.

Epidemiologists treat that as an assumption to test. An effect measured in one population is evidence about that population; extending it to a neighbouring one is an extrapolation, and extrapolations can fail.

What did ESCAPE-MeVO find?

ESCAPE-MeVO (Goyal and colleagues, N Engl J Med 2025;392:1385-1395) was a multicenter, randomized, open-label trial with blinded outcome evaluation. "Open-label" means patients and treating teams knew which treatment was given; "blinded outcome evaluation" means the people scoring recovery at 90 days did not.

The trial enrolled 530 patients from five countries between April 2022 and June 2024. All presented within 12 hours of when they were last known to be well and had favorable baseline brain imaging. Of these, 255 were assigned to thrombectomy plus usual care and 275 to usual care alone. Most patients, 84.7 percent, had the blockage in a branch of the middle cerebral artery.

The primary outcome was the share of patients with a modified Rankin score of 0 or 1 at 90 days, meaning no symptoms or no significant disability:

• Thrombectomy: 106 of 255 patients, 41.6 percent.
• Usual care: 118 of 274 patients, 43.1 percent.
• Adjusted rate ratio 0.95, with a 95 percent confidence interval of 0.79 to 1.15 (P = 0.61).

A rate ratio of 1 would mean the two groups did equally well. The estimate here sits just below 1, and the confidence interval, the range of effects compatible with the data, runs from a meaningful disadvantage to a modest advantage. The trial did not show a benefit.

The safety findings went the other way. Mortality at 90 days was 13.3 percent with thrombectomy and 8.4 percent with usual care (adjusted hazard ratio 1.82, 95 percent confidence interval 1.06 to 3.12). Symptomatic intracranial hemorrhage, bleeding inside the skull that causes symptoms, occurred in 5.4 percent of the thrombectomy group (14 of 257) and 2.2 percent of the usual-care group (6 of 272). The trial was funded by the Canadian Institutes of Health Research and by Medtronic, a device manufacturer.

What did DISTAL find?

DISTAL (Psychogios and colleagues, N Engl J Med 2025;392:1374-1384) asked the same question with a wider net. It randomly assigned participants within 24 hours of when they were last seen well, rather than 12, and it covered the broader set of medium and distal segments listed above.

The analysis included 543 participants: 271 assigned to thrombectomy plus best medical treatment and 272 to best medical treatment alone. They were older than many stroke trial populations, with a median age of 77, and 44 percent were women. Their strokes were relatively mild on arrival: the median score on the National Institutes of Health Stroke Scale, which runs from 0 to 42 with higher scores meaning more severe symptoms, was 6. Intravenous thrombolysis, clot-dissolving medication given through a vein, was given to 65.4 percent of participants. The most common blockage sites were the M2 segment (44.0 percent), the M3 segment (26.9 percent), the P2 segment (13.4 percent) and the P1 segment (5.5 percent).

DISTAL analysed its primary outcome differently from ESCAPE-MeVO. Instead of splitting patients into "good outcome" and "not," it used a shift analysis, comparing the whole distribution of modified Rankin scores between the groups. The result is a common odds ratio, where a value above 1 would favor thrombectomy:

• Common odds ratio for improvement 0.90, 95 percent confidence interval 0.67 to 1.22 (P = 0.50).

The authors reported no significant difference. All-cause mortality was 15.5 percent with thrombectomy and 14.0 percent without, and symptomatic intracranial hemorrhage was 5.9 percent and 2.6 percent. The authors described both as similar between the groups. The trial was funded by the Swiss National Science Foundation and others.

Why might a proven treatment not carry over?

The papers report what happened, not a settled explanation of why. What follows is general epidemiological reasoning, offered as questions the results raise rather than conclusions they prove.

The size of the possible benefit. A downstream branch supplies a smaller region of brain than a main artery. Less tissue at risk can mean less to gain from restoring flow, while the procedure itself still carries risk. When the possible benefit shrinks and the risk does not shrink with it, the balance can tip.

How well the comparison group did. In ESCAPE-MeVO, 43.1 percent of patients receiving usual care reached a score of 0 or 1. When many patients in the control arm already do well, there is less room for a treatment to show an improvement.

Who was enrolled. DISTAL's participants had relatively mild deficits on arrival, and about two thirds received thrombolysis. Both features shape what a comparison can detect.

The broader lesson does not depend on which explanation is right. Thrombectomy's benefit for large-vessel occlusion was established in that population. These trials show why the same benefit could not be assumed one branch further down, and why the question needed its own randomized evidence. The contrast with thrombectomy for large-core stroke, where randomized trials overturned an old assumption in the other direction, makes the same point from the opposite side: the population defines the answer.

What the trials do not settle

Negative trials have limits too.

• They describe the patients they enrolled. Each trial used specific time windows, imaging criteria and vessel definitions. The findings speak to people who resemble those participants.
• Averages are not individuals. A trial reports an average effect across a group. It does not predict the result for any single person.
• Subgroups were not the question. Neither trial was designed to answer narrower questions, such as whether a particular vessel segment or a particular kind of patient might respond differently. Exploring those is a matter for further research, not for reading between the lines of these results.
• Open-label treatment. In ESCAPE-MeVO, treating teams knew the assignment, which the blinded outcome assessment is designed to offset but cannot fully remove as a consideration.

Frequently asked questions

What is a shift analysis? A comparison of the full spread of scores between two groups, rather than a split at a single cut-off.

Why can a negative trial still be informative? Because it tests an assumption directly. Before these trials, the benefit in smaller vessels was a reasonable guess from larger ones. Now there is randomized evidence that, in the populations studied, the guess did not hold.

Do these results change what is known about large-vessel stroke? No. Neither trial studied large-vessel occlusion. Both papers describe thrombectomy for large-vessel stroke as established, and their results concern medium and distal vessels only.

A bounded, useful result

ESCAPE-MeVO and DISTAL are a clean example of why evidence is tied to the population it comes from. A treatment with a large, well-documented effect in one setting was tested in a neighbouring one and, in two independent trials, did not improve recovery at 90 days. It marks one edge of that evidence without diminishing what thrombectomy achieves where it works. Readers who want to follow stroke outcomes research into the primary literature can continue with the peer-reviewed publications.