What this is

A phase 2a dose-escalation safety trial of JX10 (formerly TMS-007), a novel dual-mechanism small molecule — promoting physiological fibrinolysis by inducing a plasminogen conformational change, plus soluble epoxide hydrolase inhibition for anti-inflammatory effect — administered as a single IV infusion to Japanese acute ischemic stroke patients 5–12 hours from last known normal. The primary safety endpoint (symptomatic ICH within 24 h with NIHSS worsening ≥ 4) was met (0% JX10 vs 2.6% placebo), and a day-90 mRS 0–1 signal (40.4% vs 18.4%, P = 0.03) emerged as a hypothesis-generating efficacy finding. Phase 3 efficacy design is the open question.

Study

Niizuma K, Nishimura N, Hasegawa K, et al. Anti-Inflammatory Thrombolytic JX10 (TMS-007) in Late Presentation of Acute Ischemic Stroke. Stroke. 2024;55:2786–2795. DOI: 10.1161/STROKEAHA.124.048464. PMID: 39508107 / PMC11593998.

1. Shadow Audit

The most underweighted number in this paper is the 0% upper bound of the 95% CI for sICH (0.0–5.6%) combined with the 22-percentage-point mRS 0–1 gap at day 90 (40.4% vs 18.4%, P = 0.03). The standard phase 2a framing treats both as exploratory — and they are. But the combination of a clean safety signal plus a clinically meaningful functional signal in the same small dataset is exactly what phase 2a is designed to detect, and the standard “hypothesis-generating only” language tends to obscure how rare that combination is in late-window stroke trials. TIMELESS (tenecteplase 4.5–24 h) had a similar design and returned a flat null on functional outcome; CHOICE (intra-arterial alteplase post-thrombectomy) had a hint of efficacy but no clean safety margin. JX10’s signal sits in a narrow zone that warrants phase 3 follow-up specifically, not general late-window skepticism.

2. Inversion Engine

For the opposite conclusion — that JX10 is harmful or non-efficacious — you would need either (a) a phase 3 trial that fails to reproduce the mRS 0–1 signal, OR (b) accumulation of sICH events in expanded dosing that the phase 2a sample size could not detect. The phase 2a sample-size constraint means a single sICH event in the JX10 arm across a 200-patient phase 3 would push the safety profile back into equipoise with tenecteplase/alteplase; the dose-response curve has not been characterized enough to know whether the 6 mg/kg cohort (the largest, n = 28) is on the plateau of the efficacy curve or still climbing into the bleeding-risk zone. The 1 mg/kg cohort’s all-male composition (0% female) is also a small-sample artifact that a larger trial will redistribute.

3. Second-Order Catalyst

If the day-90 mRS signal holds in phase 3, JX10 would become the first fibrinolytic specifically indicated for the 5–12 hour late window — the population currently routed to tenecteplase-or-nothing decisions in centers without advanced imaging, or to thrombectomy-or-medical-management in drip-and-ship networks. The clinical pathway impact is asymmetric: every late-presenting LVO patient at a non-thrombectomy center without CTP access is currently denied recanalization therapy. JX10 would not need advanced imaging to be used, which is structurally different from the WAKE-UP/TIMELESS/EXTEND imaging-selection paradigm. The pharma implication: a late-window thrombolytic without imaging selection requirements is a much larger addressable population than any imaging-selected late-window drug.

4. Asymmetric Leverage

The asymmetric payoff is the dual mechanism — fibrinolysis via plasminogen conformational change plus anti-inflammatory soluble epoxide hydrolase inhibition. If the day-90 signal is mediated by either mechanism alone, future trial design can isolate the active component and potentially develop a smaller, cheaper, more scalable molecule. The 2 mg/kg tenecteplase-vs-JX10 head-to-head (which is the natural phase 3 comparator) would also clarify whether the mechanism matters at all or whether any late-window fibrinolytic activity above the alteplase ceiling produces the same signal. The leverage is in the mechanistic disambiguation, not just the binary efficacy question.

5. Paradigm Destroyer

What does this paper kill? It kills the assumption that late-window thrombolysis requires advanced neuroimaging selection. Every prior late-window trial (WAKE-UP, TIMELESS, EXTEND, DEFUSE-3) used imaging mismatch to enrich for salvageable tissue. JX10 enrolled 5–12 hours from LKN without mandated imaging selection — and produced a meaningful functional signal in a phase 2a. That does not yet overturn the imaging paradigm (phase 3 will), but it removes the strongest argument against testing non-imaging-selected late-window fibrinolysis. The shift is from “late-window thrombolysis is only for the imaging-eligible” to “any late-presenting patient without hemorrhage on CT is potentially eligible.”

MVP — Minimum Viable Proof

For the bedside clinician: JX10 is investigational and not yet available outside trials. The phase 2a signal is encouraging but not practice-changing. Continue standard late-window decision-making: alteplase within 4.5 h, tenecteplase within 4.5 h (or per local extended protocol), thrombectomy for LVO within 24 h with imaging selection. Watch for phase 3 results.

Best Combination

Pair this with the parallel late-window pipeline: TIMELESS (tenecteplase 4.5–24 h, imaging-selected, neutral), WAKE-UP (alteplase 4.5–9 h with DWI-FLAIR mismatch, positive), EXTEND (alteplase 4.5–9 h with CTP mismatch, positive but small), and CHOICE (intra-arterial alteplase post-thrombectomy, positive). JX10 sits at a different point in the design space (no imaging selection required, 5–12 h window, dual mechanism). The optimistic reading: a portfolio of late-window options stratified by center capability (CTP access, thrombectomy access, time from onset) is emerging. The skeptical reading: most phase 2a signals in stroke do not survive phase 3 (the trial graveyard is well-populated).

Overvalue Warning

Three things to not over-read. First, the mRS 0–1 P value of 0.03 is from a small, sequential-cohort, dose-escalation design — exactly the setup where sampling variability is large and P-values should be interpreted as a screening signal, not a confirmatory result. Second, the 0% sICH figure is from 52 patients; a 1% sICH event rate in a phase 3 of 200 patients (i.e., 2 events) would still be compatible with the phase 2a point estimate but would change the safety calculus materially. Third, the Japanese-only population limits generalizability to non-Asian populations where bleeding risks and comorbidity profiles differ; a global phase 3 would need stratified analysis to detect ethnicity-specific effects.