It has apparently become time for the mobile stroke unit to go from niche to prime-time. Previously consigned to the pages of Stroke and similar journals, the latest and most comprehensive trial now graces the pages of the New England Journal of Medicine.
A flagship study for a flagship journal. Not to mention, a positive study – almost too good to be true.
Which, of course, means it probably isn’t.
In brief, the BEST-MSU trial presented is a series of nested protocols and substudies under the umbrella of purpose to test and refine the deployment of a mobile stroke unit. Amongst its many substudies are those testing time-to-tPA administration, early administration of reversal agents for intracerebral hemorrhage, and whether an on-board neurologist can be replaced by a telemedicine neurologist. The trial design is one of those “one week on, one week off” non-randomized designs – not truly randomized, but generally hoping these temporal controls generate basically similar patient populations. The primary outcome is “utility weighted modified Rankin Scale score” at 90 days, a probably-unnecessary rejiggering of what amounts to chasing a mRS 0-1.
The clear and obvious winner: mobile stroke units, the manufacturer of which two authors are consultants. In fact, MSUs were profoundly better – 55% of modified intention-to-treat patients ultimately achieved mRS 0-1 during MSU deployment weeks, compared with only 44% of those treated during traditional EMS response weeks. This treatment effect was basically driven by an 11% absolute excess of patients achieving mRS of 0.
The dramatic enhancement of treatment provided by MSU? 97.5% of those in the MSU cohort received tPA, compared with a mere 80% in the EMS cohort – and those receiving tPA did so within 60 minutes of symptom onset for 30% of those in the MSU cohort, compared with 2.6% of those with EMS. Proponents of MSUs would point to this dramatic improvement in outcomes to be associated with such “hyperacute” stroke treatment.
And, effectively, that is the principle upon which interpretation of this trial hinges. Nearly 17% of patients in the MSU cohort analysis received a final diagnosis of “stroke reversed by tPA”, as compared to 9% in the EMS cohort. There were a few more patients in the MSU cohort whose pre-existing mRS scores were 0, as well as a small excess of NIHSS 0-5, but it’s fairly clear most of the effect size on the primary outcome is driven by those final diagnoses of “stroke reversed by tPA”. This concept of “stroke reversed by tPA” is not specifically defined anywhere in the protocol or paper, somewhat limiting appraisal, unfortunately.
The traditional definition used by neurologists for “stroke reversed by tPA”, also sometimes known as “aborted stroke”, is usually equivalent to “neuroimaging negative cerebral ischemia”. NNCI occurs when patients have resolution of clinical symptoms of stroke, yet follow-up MRI shows no diffusion-weighted imaging lesions. The tPA, therefore, has saved all the brain and there is no indication there was ever any injury. However, this concept is controversial, and both emergency physicians and neurologists believe this to be relatively rare – and that these NNCI are in fact stroke mimics whose presence in study cohorts systematically bias the outcomes.
Another clue regarding those “stroke reversed by tPA” patients comes from the BEST-MSU substudy evaluating the feasibility of teleneurology versus on-board vascular neurologist. When the vascular neurologist was on-board, only 4% were “stroke reversed by tPA”, along with 10% stroke mimics. With the teleneurologist, 26% became “stroke reversed by tPA”, along with 12% stroke mimics. The sample sizes in this substudy were small, but it does not take much imagination to hypothesize there is a diagnostic accuracy component driving the “stroke reversed by tPA” final diagnoses.
There are also various issues with their CONSORT diagram. There were 10,443 stroke alerts, with 8,928 excluded as non-stroke, stroke with exclusions, or stroke mimics. Then, even though the time periods for enrollment were equal, the MSU cohort accumulated 886 enrollments versus 629 in the EMS cohort. This excess in enrollment accumulation in the MSU group is curious, and raises suspicions regarding the presence of a systematic bias towards study enrollment and treatment during MSU deployment weeks. The study protocol then trims down these enrollment cohorts into the modified-intention-to-treat population with a retrospective adjudication of tPA eligibility, further reducing the size of each cohort by approximately 30% – many of whom actually did receive tPA. This unbalanced, retrospectively tweaked cohort represents the mITT population for their primary analysis.
So, the final perspective on the utility of these multi-million dollar technological marvels comes down to, after minor whinging about their study population, the values in Table 5 of their Supplementary Appendix: did early tPA “reverse” strokes – or did their enthusiasm for MSU bias their cohort towards early treatment of stroke mimics?
“Prospective, Multicenter, Controlled Trial of Mobile Stroke Units”
https://www.nejm.org/doi/full/10.1056/NEJMoa2103879
Thanks for breaking this down thoughtfully (as usual). The tPA literature (and its offshoots) are so blatantly fraught with bias that it makes being on the delivery end of these treatments truly uncomfortable. Never mind that the B_PROUD 2.0 trial showed significantly less treatment ‘effect’ nor that the cost-effectiveness of deploying these units on a population scale have not really been analyzed, it would seem the train—erm MSU I guess—has already left the station.
Alas, in a world where “time is brain” is still somehow the best we can come up with to capture the nuance of acute stroke care it surely wont be long before ED docs are moonlighting on MSU buses.