On January 30, 2026, The Lancet published results from the EMPHASIS trial — a multicenter, double-blind, randomized, placebo-controlled study of oral minocycline in 1,724 patients with acute ischaemic stroke across 58 hospitals in China — reporting a modest improvement in 90-day functional outcomes with a 4.5-day minocycline course. A subsequent correspondence published in The Lancet identifies three specific design constraints that limit the trial’s immediate generalisability and that carry direct implications for how the next generation of neuroprotection protocols should be constructed.
The correspondence arrives at a consequential moment. No neuroprotective agent has received FDA approval for use after stroke, with the only pharmacological treatments approved for acute ischaemic stroke remaining thrombolytic agents — specifically tissue plasminogen activators. EMPHASIS is therefore not simply a single trial result; it is a data point in a decades-long sequence of neuroprotection failures, and the methodological critique in the correspondence has implications beyond minocycline specifically.
Three Design Constraints, Documented
The first constraint concerns patient selection. EMPHASIS enrolled patients with an NIHSS score between 4 and a defined upper bound, within 72 hours of stroke onset, with a median age of 65 and 33% women. That 72-hour enrollment window is wide by current reperfusion-era standards. Reperfusion therapy — intravenous thrombolysis and endovascular thrombectomy — operates in a far narrower biological window, and the correspondence flags that the EMPHASIS cohort almost certainly includes a substantial proportion of patients who received neither intervention. Preclinical and translational research increasingly supports the position that reperfusion is a necessary condition for neuroprotective treatments to produce a measurable effect — meaning a mixed cohort of reperfused and non-reperfused patients structurally attenuates the signal any neuroprotective agent can produce.
That attenuation is not a minor statistical nuisance. Approximately 7% of all acute ischaemic strokes are eligible for endovascular thrombectomy under AHA 2018 criteria, rising to 47% among anterior circulation large vessel occlusion strokes. If EMPHASIS enrolled proportionally, a large share of participants had no large vessel occlusion and received no reperfusion intervention — the subgroup where neuroprotection theory predicts the weakest benefit. Future protocols that restrict enrollment to confirmed reperfusion-treated patients will require explicit pre-stratification and will need to demonstrate that the enrolled population mirrors the intended treatment population at launch.
The second constraint is statistical. The correspondence raises concerns about whether the 90-day modified Rankin Scale outcome was adequately powered to detect a difference of the magnitude EMPHASIS observed, given the heterogeneity of the enrolled population. Post-hoc analyses by the HERMES Investigators have established that detecting endovascular treatment benefit at 80% power using the 90-day mRS as the primary outcome requires approximately 240 patients in more homogeneous, LVO-restricted populations. Applying that benchmark to a heterogeneous 1,724-patient cohort that includes non-LVO and non-reperfused patients implies that the effective powered subgroup may be considerably smaller than the enrolled total — and that a “modest” result in the overall population could mask a clinically meaningful signal in a more precisely defined subpopulation, or alternatively could be a statistical artifact of insufficient stratification.
This matters operationally. Sponsors designing Phase 3 neuroprotection trials that intend to use the 90-day mRS as a primary endpoint — the conventional functional outcome measure — will need to pre-specify the analytic population with precision, including whether the primary analysis is the ITT population, the per-protocol population, or a reperfusion-treated subgroup. An FDA Type B meeting request for a neuroprotection IND in 2026 that does not address this stratification question directly will almost certainly surface it in agency feedback.
The Generalisability Problem Is Structural
The third constraint is geographic and demographic. EMPHASIS was conducted exclusively across 58 hospitals in China. The correspondence identifies generalisability as a limiting factor — and the structural basis for that concern is well-documented. A published analysis of acute ischaemic stroke trial recruitment found that among 1,256 patients with acute ischaemic stroke in general practice, only 77 patients — roughly 6% — would have been eligible for the representative stroke trials evaluated. Geographic restriction compounds that selectivity: stroke epidemiology, comorbidity burden, time-to-treatment infrastructure, and reperfusion access rates differ materially between the Chinese hospital network EMPHASIS used and the Western European or North American settings where a regulatory approval would need to demonstrate relevance.
That gap has direct regulatory consequences. The FDA’s guidance on extrapolation of foreign clinical data, consistent with ICH E5, requires that sponsors submitting data from non-US trials demonstrate that the patient population, disease definition, concomitant therapies, and endpoint definitions are sufficiently similar to the intended US population to support the proposed labeling. For a neuroprotection trial conducted entirely in China, a sponsor seeking FDA approval would need to address — at minimum — differences in reperfusion therapy penetration rates, time-to-treatment distributions, and stroke subtype frequencies. A bridging study or US-based cohort sub-study would likely be required before FDA would accept EMPHASIS-scale data as the primary efficacy basis for a US NDA.
The pattern EMPHASIS sits within is consistent. The trial was designed as a multicenter, double-blind, randomized, placebo-controlled study — methodologically rigorous by conventional standards — yet the correspondence identifies constraints that have characterized the neuroprotection literature for more than two decades. The failure of prior neuroprotective strategies in broad clinical trials, despite promising preclinical data, has been attributed in part to precisely the population heterogeneity and reperfusion mismatch issues the EMPHASIS correspondence now raises again.
What Sponsors Must Specify This Quarter
Sponsors with active or planned neuroprotection INDs should treat the EMPHASIS correspondence as a protocol stress-test checklist. Three specifications require immediate attention before the next FDA interaction. First, enrollment criteria must define reperfusion status explicitly — not as a stratification variable post-hoc, but as a primary eligibility criterion or a pre-specified primary analysis stratum. Second, the statistical analysis plan must pre-specify the primary analytic population and demonstrate that the 90-day mRS is adequately powered within that population, referencing the HERMES benchmark of 240 patients at 80% power for LVO-restricted cohorts as a lower bound for sample size rationale. Third, for any program relying on data from non-US sites as the primary efficacy evidence, a written plan for demonstrating population comparability under ICH E5 principles should be prepared before the Type B meeting — not in response to an agency information request after filing.
The next signal to watch is whether the EMPHASIS investigators publish a pre-specified subgroup analysis stratified by reperfusion treatment status. If that data exists within the trial dataset, its presentation — or absence — at a major stroke neurology conference in the second half of 2026 will clarify whether the modest overall result reflects a diluted treatment effect in a heterogeneous population or a genuine ceiling on what minocycline can achieve even in optimally treated patients. That distinction will determine whether a next-generation protocol built around reperfusion-restricted enrollment is worth pursuing.
References
- The Lancet — “Minocycline for acute ischaemic stroke: limitations of the EMPHASIS trial” (Correspondence, 2026)
- PubMed — EMPHASIS trial: efficacy and safety of minocycline in acute ischaemic stroke, patient demographics and enrollment criteria
- American College of Cardiology — EMPHASIS trial journal scan, February 2026
- Frontiers in Neurology — “Neuroprotective strategies in acute ischaemic stroke: reperfusion as a necessary condition,” 2022
- PMC — FDA-approved treatments for acute ischaemic stroke; absence of approved neuroprotective agents
- Journal of the American Heart Association — HERMES Investigators post-hoc analysis: sample size requirements for 90-day mRS at 80% power
- PubMed — Endovascular thrombectomy eligibility rates: 7% of all AIS, 47% of anterior circulation LVO strokes (AHA 2018 criteria)
- PMC — Recruitment of ischaemic stroke patients in clinical trials: generalisability analysis, 2010
Moe Alsumidaie, MBA, MSF, is founder and Chief Editor of Vanguard Publications, which publishes Clinical Trial Vanguard, Pharma Vanguard and BullScope, and Head of Research at CliniBiz. He has two decades in clinical trial operations and data science, with earlier roles at Genentech, Abbott Vascular and Stanford University Medical Center, and is a guest lecturer in clinical trial sciences at Rutgers University.
