Comment
Continued hope for late neuroprotection with minocycline
Michael D Hill
Lancet Available online 30 January 2026
https://doi.org/10.1016/S0140-6736(26)00047-4
Neuroprotection in ischaemic stroke commonly means a treatment that results in the preservation of brain tissue in the setting of either focal or global ischaemia. The nominal focus on neuronal tissue—the neuro in neuroprotection—might be misplaced, as the ultimate goal is preservation of the entirety of the relevant tissue, including neurons, glia, vascular elements, and structural tissue. Cytoprotection of the neurovascular unit1 could be a better term.2 Neuronal intolerance to ischaemia is highly time dependent, and mechanisms to protect neurons could also protect and salvage other cellular elements that remain viable after longer durations of ischaemia. Neuroprotection research and experimental models initially focused on acute ischaemic stroke in the first hours after onset; however, evidence has shown that ischaemic injury continues to progress over hours and days. Therefore, timepoints and targets in the ischaemic process are important considerations when discussing neuroprotection.
Yao Lu and colleagues3 conducted a randomised controlled trial examining minocycline, the tetracycline antibiotic, as a neuroprotective agent, postulating that it could affect post-stroke inflammatory processes. Minocycline was first introduced in the early 1970s and has been most prominently used for the treatment of acne and rheumatoid arthritis, after being superseded by tetracycline and doxycycline as an antibiotic. The drug has been shown to be disease-modifying in clinically isolated syndromes in multiple sclerosis,4 and has been proposed to reduce intracranial haemorrhage due to its inhibition of matrix metalloproteinase-9.5, 6, 7
In the EMPHASIS study,3 1724 adult patients with ischaemic stroke were randomly assigned within 72 h of stroke onset to receive minocycline, beginning with a loading dose of 200 mg followed by a maintenance dose of 100 mg every 12 h for the next 4 days. Patients had a median age of 65 years (IQR 57–71; maximum age 80 years as per protocol) and women were under-represented (female:male 1:3). All patients were Chinese. Hypertension was common in two-thirds of patients, and diabetes in a third, whereas atrial fibrillation was uncommon (5%). The major adjudicated causes of stroke were large-artery atherosclerosis (45%) and small vessel lacunar disease (40%). Stroke was mild, with a median National Institutes of Health Stroke Scale (NIHSS) score of 5 (4–7) at randomisation. Most importantly, randomisation occurred on day 2, at a median of 41 h (26–52) after stroke onset.
Patients were randomly assigned using a permuted block design, stratified by site and with a fixed block size of four. This approach provides perfect balance between groups, but a third of assignments will be deterministic. Nevertheless, there was excellent balance on measured baseline factors. The primary outcome, defined as excellent functional neurological outcome on the modified Rankin Scale score (mRS 0–1), was conducted on a modified intention-to-treat population, which included patients who received at least one dose of study drug. Therefore, 23 randomly assigned patients were excluded from the primary analysis.
More patients (34 patients; a risk difference of 5·2%) in the minocycline group had an excellent functional neurological outcome (mRS 0–1) compared with placebo (adjusted risk ratio 1·11, 95% CI 1·03–1·20). This result was consistent in both the per-protocol analysis and after multiple imputations to account for missing data in the randomly assigned population. However, the result was sensitive to more extreme assumptions (ie, a worst-case imputation scenario in which all missing cases were considered unfavourable for minocycline and favourable for placebo) and to tipping point analysis. Whether a 1-point greater average improvement in the NIHSS score at 6 days, and if 1·2% fewer patients suffering neurological worsening at 6 days, is consistent enough with a 5·2% 90-day benefit on the modified Rankin Scale score, is unclear. The absence of clear improvement on other secondary outcomes means that corroborative internal evidence from the trial is not available.
No clear subgroups particularly benefited from minocycline. Stroke of undetermined aetiology seemed to show a greater benefit with minocycline but the biological implausibility of this observation suggests that this is a random finding. Future trials could investigate the effect of time on minocycline efficacy, as greater benefit with earlier treatment is biologically plausible.
Overall, the possibility of having novel adjuvant agents for stroke beyond reperfusion is tantalisingly close. Lu and colleagues should be congratulated for their work, as the trial provides randomised evidence favouring the validity of the target of the inflammatory processes that occur in the hours to days after acute stroke. More trials of minocycline or other agents that target these processes are needed. Internally consistent evidence on several outcomes—in addition to the chosen primary outcome—is also required, including interim imaging biomarkers such as infarct volume or reduction in oedema, and blood biomarkers that are known to correlate with the degree of tissue injury, such as S100β, neurofilament light chain, and brain-derived Tau.8 In addition, true randomness of allocation is an important principle upon which statistical inference relies, and small block sizes should be abandoned when permuted block randomisation approaches are used. Consistency of effect in future trials will help us understand if the trial's result can be relied upon. Because minocycline is cheap, generic, and widely available, it has immediately become a candidate for future trials. It could be assessed again in patients with more severe stroke, perhaps with an intravenous formulation to manage the problem of drug administration in patients with dysphagia. For now, this result will be important for researchers in the stroke community, as more evidence to support and corroborate the EMPHASIS trial is now needed.
