Mobile stroke units: what the trials show
Mobile stroke units bring a CT scanner to the patient. What the trials found, how they were designed without classic randomization, and what is unsettled.
By Manouchehr Hessabi, MD, MPH
A mobile stroke unit is an ambulance that carries a computed tomography (CT) scanner, a point-of-care laboratory, and stroke expertise, so that the clot-dissolving drug used in many ischemic strokes can be started at the curb instead of after arrival at the hospital. In the largest controlled studies, these units shortened the time to treatment by roughly half an hour and were associated with less disability at 90 days. A later study period in Berlin did not show a statistically significant benefit on its own, and questions about cost and staffing remain open.
The idea follows from a well-established finding, restated by the Berlin investigators discussed below: the effects of thrombolysis, treatment with a drug that dissolves the clot blocking an artery, are time-dependent. The drug used in these trials is tissue plasminogen activator, abbreviated t-PA. The open question is whether a costly ambulance delivers enough extra speed, to enough people, to change how they recover.
How do researchers measure the outcome?
Most of the studies below use the modified Rankin scale, a seven-level rating of disability after stroke that runs from 0 (no symptoms) to 6 (death). The pooled analysis discussed below defines a score of 0 or 1 at 90 days as an excellent outcome.
Some analyses use a single cut point, such as the share scoring 0 or 1. Others ask whether one group has been shifted toward less disability across every level, summarized as a common odds ratio. Its direction depends on how the authors defined it, which matters in the Berlin studies below.
Why is prehospital stroke treatment hard to study?
In a randomized controlled trial, patients are assigned by chance, so the groups differ on average only in treatment. That protects against confounding, where another factor that differs between groups produces what looks like a treatment effect.
Mobile stroke units make individual randomization awkward. The intervention begins with the emergency call, before anyone knows whether the patient has had a stroke, so the unit has to be sent or not sent before a study participant has been identified. The studies below used two workarounds.
- Allocation by time. The unit is available on some weeks or days and not on others, in a fixed or randomized schedule. Patients are grouped by which period their call fell into.
- Allocation by availability. The unit is dispatched whenever it happens to be free, and outcomes are compared between patients who had it dispatched and those who did not.
Allocation by time spreads unknown differences between patients across both arms, provided nothing else about care changes between periods. Allocation by availability depends on the unit's availability being unrelated to who is calling and where. Neither is equivalent to randomizing individual patients, and the investigators describe their own studies accordingly.
What did BEST-MSU find?
The largest U.S. study, BEST-MSU, published in the New England Journal of Medicine in 2021, is described by its authors as an observational, prospective, multicenter, alternating-week trial. Patients were managed either by a mobile stroke unit or by standard emergency medical services (EMS), depending on the week.
The study enrolled 1,515 patients, of whom 1,047 were eligible for t-PA: 617 in the mobile unit group and 430 in the EMS group.
The median time from stroke onset to t-PA was 72 minutes with the mobile unit and 108 minutes with EMS. Among eligible patients, 97.1% in the mobile unit group received t-PA, compared with 79.5% in the EMS group.
The primary outcome used a utility-weighted version of the modified Rankin scale. Utility weighting converts each disability level into a value between 0 and 1 according to a patient value system, rather than treating every step as equal. The mean utility-weighted score at 90 days was 0.72 in the mobile unit group and 0.66 in the EMS group, with an adjusted odds ratio of 2.43 (95% confidence interval 1.75 to 3.36) for reaching the threshold that approximates a Rankin score of 0 or 1. In plain terms, 55.0% of eligible patients in the mobile unit group and 44.4% in the EMS group had a score of 0 or 1 at 90 days. The abstract reports mortality at 90 days of 8.9% in the mobile unit group and 11.9% in the EMS group.
What did the Berlin studies find?
The first Berlin study, B_PROUD, published in JAMA in 2021, was a prospective, nonrandomized, controlled intervention study run from February 2017 to October 2019. When a call suggested stroke, a conventional ambulance was sent, along with a mobile stroke unit if one was available. The analysis compared 749 patients who had a unit dispatched with 794 who had a conventional ambulance only.
Patients with a unit dispatched had a lower median Rankin score at three months (1 versus 2). The common odds ratio for a worse score was 0.71 (95% confidence interval 0.58 to 0.86). In this study a value below 1 means a shift toward less disability. The authors concluded that dispatch was significantly associated with lower disability, and that clinical trials in other regions were warranted.
The follow-on study, B_PROUD-2.0, published in Neurology in 2025, ran from May 2019 to April 2021 and was also nonrandomized and controlled. The authors report that the COVID-19 pandemic and limited funding hindered full implementation of planned improvements, and the study included 1,050 patients against 1,500 planned.
On its own, B_PROUD-2.0 found no statistically significant effect on the primary outcome: a common odds ratio of 0.90, with a 95% confidence interval of 0.72 to 1.14. Treatment speed still differed sharply, with far higher odds of thrombolysis within one hour of dispatch in the mobile unit group (odds ratio 10.15). When the investigators pooled the primary populations of both Berlin study periods, 2,666 patients in all, the common odds ratio was 0.80 (95% confidence interval 0.67 to 0.96), a favorable shift.
This is a distinction central to reading any trial. The B_PROUD-2.0 interval runs from a meaningful benefit (0.72) to a slight harm (1.14), and the authors call the study underpowered: that period could not tell the two apart. That is absence of evidence from one period, not evidence that the effect disappeared, and not proof that it persisted. The abstract's formal classification states Class III evidence that adding mobile units did not improve three-month outcomes in that period. It belongs beside the pooled result, not in place of it.
What does the pooled evidence say about benefit and safety?
A systematic review and meta-analysis in JAMA Neurology in 2022 combined the studies available at the time. Compared with usual care, mobile stroke unit use was associated with excellent outcome, a Rankin score of 0 or 1 at 90 days (adjusted odds ratio 1.64; 95% confidence interval 1.27 to 2.13; 5 studies, 3,228 patients). Across 13 studies and 3,322 patients, the median reduction in time from onset to intravenous thrombolysis was 31 minutes.
On safety, mobile unit use was not associated with increased all-cause mortality at 7 or 90 days, or with a higher proportion of symptomatic intracranial hemorrhage, bleeding in the brain that causes new symptoms, after thrombolysis.
Two limits apply. Pooling mostly nonrandomized studies makes an estimate more precise but does not remove bias shared across them. And this analysis predates B_PROUD-2.0.
What remains unsettled?
Cost. A cost-effectiveness analysis in the Annals of Neurology in 2025, based on a prospective, multicenter, alternating-week trial of mobile units versus EMS and taking the perspective of Medicare, reported results as an incremental cost-effectiveness ratio: the extra cost of the mobile unit approach divided by the extra health it produces. Health was measured in quality-adjusted life years (QALYs), which count years of life weighted by their quality. The first-year ratio for all t-PA eligible patients, using total costs, was $238,873 per QALY. For patients without pre-existing disability it was $61,199. Over a lifetime horizon, the figures were $94,710 and $31,259 per QALY. The authors call cost-effectiveness borderline on total first-year costs across all eligible patients, and note that every ratio depended heavily on how many patients a unit treats each year.
Staffing. MSU-TELEMED, published in NEJM Evidence in 2026, was a randomized trial by investigators in Melbourne, Australia, in which care on a mobile unit was randomized by day to an onboard neurologist or a neurologist by telemedicine. Its primary outcome was a hierarchical composite judged by win odds: every patient in one group is compared with every patient in the other, first on safety, then on time from scene to treatment decision, then on how much of the neurologist's time went to direct patient care. Among 275 participants, the comparisons favored telemedicine (76% wins, 4% ties, 20% losses; adjusted win odds 3.5, 95% confidence interval 2.4 to 5.1). Safety events were similar (13% versus 12%). The decision itself took longer by telemedicine, a median of 19 minutes versus 13, while the neurologist's time spent directly in patient care was a median of 100% with telemedicine versus 33% onboard.
Generalizability. The large outcome studies were run in major metropolitan areas in the United States and in Berlin. How well the results transfer to other settings is an open limitation, not a finding about any particular place.
How should a reader weigh this evidence?
Taken together, the literature shows a consistent shortening of time to treatment across many studies, outcome benefits in the two largest controlled studies and in pooled analyses, one underpowered study period that could not confirm a benefit on its own, and no signal of increased mortality or symptomatic brain bleeding. Because none of the outcome studies randomized individual patients, "associated with" is the language that fits.
The open questions are practical ones: whether the cost is justified where fewer patients are treated each year, and whether the specialist needs to be in the vehicle at all. More on how stroke outcomes are measured and studied is on the stroke outcomes research page, and the peer-reviewed publications are the place to continue into the primary literature.