Intrathecal Exosome Therapy in Spinal Cord Injury

Executive Summary

Intrathecal exosome therapy is one of the most plausible biological ideas we have today in spinal cord injury; it is also one of the most aggressively overhyped commercial practices. Both sentences are true, and the distance between them is the subject of this review.

The answer in one paragraph

Mesenchymal stromal cell-derived extracellular vesicles (MSC-EVs; commonly marketed as exosomes) are a biologically coherent, mechanistically plausible and preclinically well-supported candidate therapy in spinal cord injury. They are also, as of mid-2026, a Phase I entity with a single published human study in nine patients: no control group, no licensed product anywhere in the world, an unresolved potency (efficacy measurement) assay problem that blocks conventional licensing, and, in parallel, a grey market that has produced serious adverse events documented by the FDA. The honest summary is this: intrathecal exosome therapy is one of the more promising items in a field with a long list of promising treatments that failed in Phase II/III — and the gap between what the laboratory evidence supports and what is sold to patients is one of the widest in contemporary regenerative medicine.

Key findings

Three sentences a physician can state with confidence

1. "There is a plausible biological rationale and a single small safety study." — True.
2. "There is no evidence that it restores motor function in humans." — True; this is the statement most frequently violated in commercial marketing.
3. "Receiving this product outside a registered clinical trial means exposure to an unmeasured risk and a known regulatory violation." — True for the United States, the EU, the United Kingdom, Canada, Japan and Australia.

Conclusion by clinical stage

How This Review Was Prepared — and What Was Not Used

This text is a critical review with a structured evidence appraisal; it is not a systematic review. The distinction is not cosmetic: attaching a PRISMA flow diagram to a literature containing a single uncontrolled clinical study would lend the visual authority of systematic methodology to what is in fact a case series. That is a category error.

Method

Sources were identified by searching the indexed literature, regulatory databases (FDA CBER safety notifications, warning letters), trial registries (ClinicalTrials.gov, IRCT), professional society position papers (ISEV MISEV2023) and national regulatory announcements. The order of priority was: the single published human study and its full methods; preclinical meta-analyses rather than individual rodent studies; current clinical practice guidelines (AO Spine 2024) as the standard-of-care anchor; and comparator trials reported after 2024 (NISCI, Up-LIFT).

Errors removed from the source texts

This review was built on two draft texts. The first was rigorous and explicit about its own limits; it was largely preserved, with several points updated. The second contained a number of important statements that were unverifiable or outright wrong, and these were not used. In the interest of transparency, that filtering is listed below, because the same errors circulate frequently in patient information material and marketing content.

Points updated from the first text

• The ExoPTEN timeline. The programme, announced in 2025 as launching Phase 1/2a in 2026, is according to the company's Q4 2025 and Q1 2026 reports still preclinical; IND-enabling studies are under way and the IND application is in preparation. In other words, the randomised controlled arm that constitutes the field's real decision point has not yet begun.
• ARC-EX regulatory detail. The December 2024 authorisation was a de novo classification; the November 2025 home-use expansion was a 510(k) clearance. The CE mark was obtained in September 2025, covering home use as well.
• The Turkish dimension. National epidemiology and regulatory status, absent from the original text, have been added.

Evidence grading convention

Oxford (OCEBM) levels for treatment: 1 — systematic review of randomised trials or a single randomised trial with a narrow confidence interval; 2 — cohort study or low-quality randomised trial; 3 — case-control; 4 — case series, poor cohort; 5 — mechanism-based reasoning, expert opinion, animal data. GRADE certainty is given as High, Moderate, Low or Very low; it is downgraded for risk of bias, indirectness (animal to human), imprecision (small samples), inconsistency and publication bias.

Spinal Cord Injury: Definition, Epidemiology and Turkish Data

Definition and classification

Spinal cord injury is damage to the spinal cord causing partial or complete loss of motor, sensory and/or autonomic function below the neurological level of injury. It is classified along three axes: aetiology (traumatic — falls, traffic, violence, sport; non-traumatic — degenerative, neoplastic, vascular, infectious, inflammatory), completeness (ASIA Impairment Scale A–E, according to ISNCSCI standards) and temporal stage (acute under 48 hours, subacute 48 hours – 6 months, chronic beyond 6–12 months).

Despite its well-documented insensitivity, the AIS grade remains the field's primary regulatory endpoint: each grade harbours marked neurological heterogeneity, and a change of grade may not track the function the patient cares about.

Global burden

Two epidemiological facts determine the urgency of treatment. First, while age-standardised rates are falling, the absolute burden is rising; decomposition analyses attribute the rise to demographic rather than epidemiological change (population growth and ageing). Second, the aetiological profile is shifting: in high-income countries, from high-energy trauma in young men towards low-energy falls in older adults with pre-existing cervical canal stenosis. This second group has a different injury biology (central cord syndrome, incomplete injury) and a different tolerance of risk when faced with regenerative therapy.

Turkish data

Because Turkey has no nationwide spinal cord injury registry, national figures rest on a small number of epidemiological studies and are out of date. The most frequently cited national study is a multicentre survey of cases from 1992.

Socioeconomic burden

International cost data vary enormously depending on accounting methods: in the United States the annual direct economic cost exceeds 9.7 billion dollars, with lifetime treatment costs per patient of 1.1–4.7 million dollars depending on the level of injury (higher for high cervical injuries and for injury at a young age); in Ontario the lifetime estimate per person is about 336,000 dollars; in China the average admission for traumatic injury costs about 11,500 dollars.

The cost distribution loads onto chronic care — bladder and bowel management, pressure ulcers, respiratory complications, spasticity, neuropathic pain and readmissions — rather than the index admission. This has a critical consequence for treatment appraisal: an intervention that improves autonomic function (bladder, bowel, cardiovascular stability) without changing the AIS grade may generate more health-economic and patient-reported value than one that moves the motor score. That is precisely the domain in which the single published exosome study gave its most consistent signals — and it is also the domain that regulatory endpoints anchored to the AIS capture worst.

Unmet clinical need

Spinal Cord Anatomy and Neurobiology

This section is abbreviated; its purpose is to establish the anatomical ground for the mechanistic argument in the sections that follow.

Grey matter and the autonomic substrate

The Rexed laminae I–X run as follows: I–VI dorsal (sensory processing), VII intermediate (including the intermediolateral cell column from T1 to L2 and the sympathetic preganglionic neurons), VIII–IX ventral (motor neurons, somatotopically arranged), X periependymal. The intermediolateral cell column and the sacral parasympathetic nucleus (S2–S4) are the anatomical substrate of the autonomic dysfunction that dominates the morbidity of spinal cord injury — and they are also the structure underlying the neurogenic bowel and urinary outcomes in the exosome study.

Glial biology

The astrocytic border (historically called the glial scar) should not be modelled as a pure obstacle. Recent work has shown that removing it worsens outcomes: the border limits inflammation and protects tissue. Strategies aimed at dissolving the scar entirely have therefore been replaced by strategies aimed at modulating its composition. This is precisely where the appeal of MSC-EVs lies: rather than abolishing astrogliosis, they shift the astrocyte phenotype (from A1 towards A2).

Pathophysiology: Why Does It Not Heal?

The two-phase model

Primary injury is mechanical — contusion, compression, distraction or laceration — and causes immediate axonal tearing, membrane disruption, haemorrhage and rupture of the blood–spinal cord barrier. No agent can reverse primary injury.

The secondary injury cascade unfolds over hours to months, and it is the part that is treatable: ionic dysregulation (sodium and calcium influx), glutamate excitotoxicity, oxidative stress, mitochondrial failure, neuroinflammation (neutrophils, then M1 macrophages; TNF-α, IL-1β, IL-6) and vascular ischaemia with oedema. Their common output is cell death — apoptosis, necroptosis, ferroptosis and pyroptosis. Loss of oligodendrocytes leads to demyelination of intact axons, and Wallerian degeneration to retrograde axonopathy. In the chronic phase, the astrocytic border, chondroitin sulfate proteoglycan deposition, a pericyte- and fibroblast-derived fibrotic core and cystic cavitation (syringomyelia) establish a permanent conduction block.

Time windows and the logic of treatment

Molecular detail of the key cascades

Excitotoxicity. Depolarisation leads to glutamate release, which leads to over-activation of NMDA/AMPA/kainate receptors and calcium influx; calpain and phospholipase A2 are activated; proteolysis of the cytoskeleton and release of arachidonic acid follow. Glutamate also acts as a chemotactic signal for microglia, linking excitotoxicity to inflammation.

Inflammation. Rupture of the barrier permits infiltration first by neutrophils and then by monocytes. Two functionally divergent macrophage populations dominate: pro-inflammatory (M1-like; TNF-α, IL-1β, iNOS and neurotoxicity) and reparative (M2-like; IL-4, IL-10, matrix remodelling). In the injured cord the M1 phenotype persists abnormally long — the failure of the M1-to-M2 transition is the single most targeted node in the MSC-EV literature.

Astrocyte polarisation. Injury-associated microglia secrete IL-1α, TNF and C1q, converting astrocytes to the neurotoxic A1 phenotype. MSC-EVs reduce A1 conversion by suppressing the phosphorylated NF-κB p65 subunit; this effect has been replicated by several independent groups and is one of the field's more robust findings.

Barrier integrity. TIMP2, present in MSC-EV cargo, inhibits matrix metalloproteinase-mediated degradation of tight junction proteins. Silencing with siTIMP2 abolishes the barrier-protective effect — meaning that this mechanism satisfies a necessity criterion that most vesicle mechanism claims cannot.

Modes of cell death. Apoptosis dominates in the subacute phase (BAX up, Bcl-2 down, caspase cascade). Necroptosis (RIPK1/RIPK3/MLKL) and ferroptosis (iron-dependent lipid peroxidation, GPX4 depletion — highly relevant given the iron load after haemorrhage) are increasingly recognised contributors. Pyroptosis via NLRP3/caspase-1/GSDMD links inflammasome activation to lytic death.

Why does regeneration fail in the adult central nervous system?

1. Extrinsic inhibition — myelin-associated inhibitors (Nogo-A, MAG, OMgp) signal through NgR1/p75/LINGO-1 and RhoA/ROCK; chondroitin sulfate proteoglycans signal through PTPσ and LAR.
2. Failure of the intrinsic growth state — adult central nervous system neurons suppress regeneration-associated genes; the two best-characterised brakes are PTEN/mTOR and SOCS3/JAK-STAT (the direct rationale for PTEN-targeting engineered vesicles).
3. Absence of a permissive substrate — the cystic cavity offers no scaffold for axonal elongation.
4. Failure of remyelination and target reinnervation — even an axon that successfully elongates must be remyelinated and form a functional synapse on the correct target.

Current Standard of Care

Appraising an experimental treatment requires an accurate baseline first. The 2023–2024 AO Spine and Praxis clinical practice guidelines represent the current international standard.

Acute management

Rehabilitation and restorative technologies

The Up-LIFT result deserves emphasis because it reframes the competitive landscape. In 65 participants with chronic incomplete tetraplegia across 14 centres, non-invasive transcutaneous stimulation combined with functional task practice met its primary safety and efficacy endpoints: about 90 percent of participants improved in upper limb strength or function, about 87 percent reported an improvement in quality of life, and benefit was demonstrated even in participants 34 years after injury; no treatment-related serious adverse events occurred.

Symptom management

Exosome Biology and Mechanisms of Action

Terminology: a question of precision with clinical consequences

ISEV's MISEV2023 guidance is explicit: the correct general term is extracellular vesicle; the term exosome should be reserved for vesicles shown to originate from the endosomal multivesicular body pathway — a demonstration that no therapeutic preparation actually provides.

MISEV2023 instead recommends reporting dose by at least two independent measures (for example particle count by nanoparticle tracking analysis or tunable resistive pulse sensing, together with protein or lipid content), characterising the source cell, and using dose-response together with appropriate vesicle and non-vesicle negative controls in functional studies.

Vesicle cargo

Comparison of source cells

Direct comparative evidence: a pooled preclinical meta-analysis found no statistically significant difference in efficacy at multiple time points between MSC exosomes derived from bone marrow, adipose tissue, umbilical cord and placenta; one synthesis found bone-marrow-derived preparations particularly effective in subgroup analysis. The correct reading is this: the comparison has not been performed with adequate power — the sources have not been shown to be equivalent.

Mechanisms of action

Anti-inflammatory and immunomodulatory effect (the best evidenced). MSC-EVs suppress nuclear translocation of NF-κB p65; they reduce TNF-α, IL-1α and IL-1β and increase IL-10 and IL-4. They reduce A1 astrocyte conversion and shift macrophage polarisation away from M1, in part by breaking the feedback loop between reactive oxygen species and M1 activation via MAPK–NF-κB p65 signalling. This is the mechanism most reproducibly demonstrated across independent groups.

Anti-apoptotic and neuroprotective effect. A consistent reduction in BAX and increase in Bcl-2, with consequent attenuation of caspase-3. Inhibition of TLR4/MyD88/NF-κB has been shown to mediate part of this effect.

Stabilisation of the blood–spinal cord barrier. TIMP2-dependent inhibition of matrix metalloproteinase-mediated degradation of tight junctions. This mechanism passes a necessity test (siTIMP2 abolishes the effect), which sets it apart from most vesicle mechanism claims.

Angiogenesis. Increased vessel density and VEGF signalling; systemic MSC-EV administration in a rat model supports angiogenesis alongside reductions in apoptosis and inflammation.

Axonal growth and remyelination (the weakest evidence). Increases in GAP-43, neurofilament and myelin basic protein together with oligodendrocyte progenitor differentiation have been reported. However, most studies show increased sprouting and preserved axons rather than long-distance regeneration across the lesion. In contusion models, histological regeneration is often indistinguishable from tissue preservation secondary to neuroprotection. This distinction is routinely erased in review articles and in commercial marketing, and it is the single largest source of hype in the field.

Synaptic plasticity and extracellular matrix. Increases in synaptophysin and PSD-95; increased sprouting in preserved reticulospinal and propriospinal fibres. Mechanistically this overlaps with the rationale for stimulation-based therapy and is the strongest argument for combining exosomes with neuromodulation. Restoration of the matrix metalloproteinase and TIMP balance, reduced proteoglycan deposition and altered composition of the fibrotic core have also been reported.

Mechanism–evidence quality matrix

The Science of Intrathecal Delivery

Comparison of administration routes

The unresolved intrathecal problem

PET tracking of radiolabelled engineered vesicles in primates has shown that distribution in cerebrospinal fluid depends on the route of administration: intrathecal delivery produced a meningeal distribution extending along the neuraxis as far as the skull base. Crucially, distribution in the primate was heterogeneous — unlike the homogeneous distribution seen in rodents. This is the most consequential and least discussed finding in the intrathecal vesicle literature. It has three implications:

1. Rodent cerebrospinal fluid pharmacokinetics do not scale. The mouse subarachnoid space is a small, well-mixed compartment. The human one is a long, compartmentalised, pulsatile system with marked flow heterogeneity, arachnoid trabeculation and — critically in spinal cord injury — post-traumatic arachnoiditis and adhesions that disturb distribution further.
2. Meningeal does not mean parenchymal. Vesicles distributed along the meninges must cross the pia mater and enter the cord parenchyma to reach microglia, astrocytes and neurons. Nanoparticle penetration depth in central nervous system tissue is on the order of hundreds of micrometres to a few millimetres — sufficient for the dorsal columns, questionable for the anterior horn motor neurons of a human cord roughly 1 cm in diameter.
3. This may explain the pattern of the clinical signal. That sensory improvement (dorsal, superficial) reached significance in the Phase I study while motor improvement (ventral, deep) did not is exactly what a diffusion-limited distribution model predicts. It is also exactly what the spontaneous recovery model predicts. Without a control and without biodistribution imaging the two cannot be separated.

Unknown pharmacokinetics

No human dose-ranging study of intrathecal vesicles in spinal cord injury has ever been carried out. Every dose administered today — in trials and, far more commonly, in commercial clinics — is empirical.

The procedure, as performed in the published study

1. Screening: ASIA A, age 18–60, 2 weeks to 6 months after injury.
2. Exclusion: other spinal pathology, intracranial lesion, active infection, immunological disorder, severe comorbidity.
3. Baseline assessment: ISNCSCI/ASIA, SCIM-III, neurogenic bowel dysfunction score, Modified Ashworth, urinary questionnaire, laboratory tests, ECG, magnetic resonance imaging.
4. Administration: under sterile conditions in theatre, lumbar puncture at L4/L5 with a 24G spinal needle; entry into the arachnoid space confirmed by free flow of cerebrospinal fluid; slow injection of 300 µg of total vesicle protein in 5 mL of phosphate-buffered saline (60 µg/mL); the needle left in place for 60 seconds (reducing the risk of leak).
5. Follow-up: weekly for 4 weeks, then monthly safety monitoring for 3 months with CTCAE v5.0 grading; imaging at 6 months; efficacy assessment at 6 and 12 months; standard rehabilitation throughout.

The safety profile of the route itself

Preclinical Evidence and Critical Appraisal

Pooled effect estimates

What do these numbers actually mean?

The Basso–Beattie–Bresnahan (BBB) scale runs from 0 to 21. From a typical post-contusion baseline of 7–9 points, a gain of 1.5–3.5 points means movement in the range of occasional to consistent weight-supported plantar steps with inconsistent forelimb–hindlimb coordination. In a rat this is a real, reproducible and biologically meaningful improvement.

It is not walking. It is not recovery. And the track record of translating BBB gains of this magnitude into human AIS conversion is historically zero.

Systematic weaknesses of the preclinical corpus

Clinical Evidence: The Entirety of the Human Data

The human evidence base for intrathecal exosomes in spinal cord injury consists of a single study. A full appraisal is set out below.

Akhlaghpasand et al. (2024) — Stem Cell Research & Therapy 15:264

Results — safety (primary endpoint: met)

A total of 7 adverse events in 5 of 9 patients (55.6 percent): 5 Grade I (71.4 percent) and 2 Grade II (28.6 percent). No adverse event related to the intervention — 4 unrelated and 3 unlikely to be related. No serious adverse events. No anatomical complication was reported on imaging at 6 months.

Results — efficacy (secondary, exploratory, underpowered)

Strengths of the study

• A first-in-human application; a genuine regulatory and scientific milestone.
• Rigorous product characterisation by the standards of the clinical vesicle field (flow cytometry, dynamic light scattering, electron microscopy, Western blot).
• CONSORT-compliant reporting; prospective registration.
• Adverse event appraisal with CTCAE grading by an independent safety board.
• Standardised surgical management (all patients decompressed within 36 hours) and standard rehabilitation — reducing between-patient variance in concomitant intervention.
• Twelve-month follow-up and safety imaging.
• The authors themselves state the efficacy limitation explicitly, including the spontaneous recovery confounder — an intellectual honesty absent from most secondary citations of this study.

Limitations — those that are fatal to any efficacy inference

The spontaneous recovery confounder, in numbers

This is the crux of the entire clinical evidence question, and therefore deserves to be stated numerically.

In a cohort of 9 ASIA A patients, between 1.5 and 2.7 patients would be expected to change grade spontaneously. Some ASIA motor gain was seen in four of the nine patients. A shift of a few points of sensory score on a 112-point scale sits comfortably within the noise and natural history envelope of a subacute cohort.

Adjacent clinical evidence: intrathecal cell therapy

Because intrathecal cell therapy is frequently confused with vesicle therapy in patient information material, the closest comparators are given below.

The pattern is both striking and instructive: uncontrolled intrathecal cell and vesicle studies in subacute spinal cord injury consistently report improvement. So does the natural history. The one design feature that could distinguish between them — randomisation — is almost entirely absent from this literature.

Ongoing and announced trials

The wider vesicle clinical landscape, for calibration

An analysis of ClinicalTrials.gov through the end of 2025 identified 355 registered trials of human-derived vesicles across all indications. Among registered MSC-EV trials, intravenous infusion and aerosol inhalation dominate; intrathecal administration is rare; and spinal cord injury constitutes a negligible fraction. The field is broad but shallow: the overwhelming majority are Phase I, small and single-arm.

Evidence grading summary

Comparison with Other Regenerative and Restorative Therapies

Exosomes or stem cells? The central strategic comparison

Exosomes and neural stem cells

Neural stem cells have a higher theoretical ceiling (they can replace lost neurons and build relay circuits) but a worse practical record. Vesicles have a lower ceiling and an easier path. Neither has demonstrated clinical efficacy. The most defensible reading is that they address different problems and are more likely to be complementary than competing.

Route comparison for exosomes

Engineered Exosomes and the ExoPTEN Programme

Engineering strategies

The leading engineered asset: ExoPTEN

Engineered or native?

Pooled preclinical data favour engineered vesicles (standardised mean difference 4.21 for miRNA-modified vesicles; 95 percent confidence interval 3.39–5.04). But funnel plot asymmetry was significant and trim-and-fill imputed four unpublished studies; the true superiority is therefore smaller than reported. Engineering also brings costs the preclinical literature does not measure: additional chemistry-manufacturing-controls characterisation, variability in loading efficiency, the possible immunogenicity of displayed peptides and a markedly longer regulatory path.

Combination Therapies

Monotherapy is unlikely to be sufficient, because the four barriers to central nervous system regeneration are independent of one another.

Manufacturing, the Potency Problem and Regulatory Status

The potency assay problem — the field's binding constraint

To license a biologic, the applicant must define a potency assay: a quantitative measure of the biological activity responsible for the clinical effect, used for batch release. For MSC-EVs this requires answering the question of what exactly the active substance is. The candidate answer is hundreds of miRNAs, proteins and lipids acting pleiotropically — which is not a licensable answer.

MISEV2023 encourages multi-metric reporting and mandatory dose-response with vesicle and non-vesicle negative controls; but MISEV is a research reporting standard, not a pharmacopoeial one. The gap between MISEV compliance and regulatory-grade chemistry-manufacturing-controls is why no exosome product has been licensed anywhere.

Manufacturing methods

Additional constraints: the requirement for serum-free or vesicle-depleted media, bioreactor optimisation, donor-to-donor variability in the source MSC line, senescence-related potency decline with passage number, and cold chain logistics.

Regulatory status

Harms from unlicensed products documented by the FDA include reports of severe infections, allergic reactions and tumour formation; across the wider regenerative medicine categories there are also reports of blindness and death.

The situation in Turkey

In Turkey, exosome products are handled by the Turkish Medicines and Medical Devices Agency within the framework for advanced therapy medicinal products. A notable development occurred in this field in 2026: the Betül-Ziya Eren Genome and Stem Cell Centre at Erciyes University completed the agency's inspection processes and received a Manufacturing Site Authorisation Certificate for the production of umbilical cord-derived mesenchymal stem cell exosomes for use in clinical applications, becoming the only public institution in Turkey holding a manufacturing site authorisation for stem cell exosomes. The assessment covered manufacturing areas, the quality management system, cleanroom infrastructure to good manufacturing practice standards, quality control processes and technical equipment.

What would a credible licensing path look like?

1. Definition of an active substance hypothesis (for example a specific miRNA or protein).
2. Development of a quantitative potency assay correlating with in vivo effect — missing, and the binding constraint.
3. Locking down the manufacturing process (tangential flow filtration with size exclusion chromatography, GMP, serum-free).
4. Demonstration of multi-metric batch comparability according to MISEV2023.
5. Large animal (pig or primate contusion model) efficacy and biodistribution — missing.
6. An IND or clinical trial application with a defined dose, route and schedule.
7. Phase I: dose escalation with pharmacokinetic and biodistribution imaging and cerebrospinal fluid biomarkers.
8. Phase II: randomised, double-blind, sham-controlled, stratified by AIS grade and level.
9. Phase III: multicentre, with a primary endpoint that is meaningful to patients.
10. Licensure and a long-term registry.

Two of these ten steps are red, and they are precisely the two things the field has not done: a potency assay and large animal efficacy data with biodistribution. Neither Phase I nor Phase II can substitute for them.

Critical Discussion

Quality of evidence: an honest summary

The intrathecal exosome literature exhibits a characteristic and recognisable structure:

• A broad, positive and biased preclinical base (funnel asymmetry in several independent meta-analyses; Egger p = 0.00 in the largest).
• A single, small, uncontrolled human study — whose limitations are stated plainly by its own authors.
• A secondary literature — reviews, commentaries, conference material, corporate communications — that strips those limitations a little further with each citation generation.
• A commercial layer — far ahead of both and operating largely independently of them.

This structure is not peculiar to exosomes. It is the same structure that preceded the collapse of methylprednisolone as a universal standard, GM-1 ganglioside, minocycline and the primary endpoint of NISCI. The field of spinal cord injury has a well-documented base rate for this pattern, and that rate is not favourable.

Citation drift: a documented and verifiable problem

Follow the claim. The primary source says: this Phase I study was not powered to assess efficacy; recovery can occur naturally in the first one to two years after injury; this limits our inferences. Secondary reviews convey it as: it was safe and well tolerated, with significant improvements in neurological and functional status at 12 months. Marketing turns it into proof that exosomes improve function after spinal cord injury.

Each step is defensible on its own. The composite is a distortion. A concrete example of how commonplace this drift can be: a peer-reviewed review from 2025 describes this study as an intradermal injection in acute spinal cord injury — whereas the study is intrathecal administration in subacute injury. Route of administration and stage of injury are the two variables that determine the entire logic of this treatment; that both could be misreported in a peer-reviewed review gives a sense of how far downstream texts can be trusted.

Conflicts of interest and structural bias

Reproducibility

The reproducibility problem in this field is primarily material rather than statistical. Two laboratories following the same published protocol will produce different products, because: the source MSC donor, passage number and culture conditions alter the secretome; the isolation method alters vesicle subpopulation composition and non-vesicular particle contamination; a dose expressed as total protein does not fix particle count; storage and freeze-thaw history alter integrity; and no reference material or international standard exists.

Open safety questions

Ethical issues

1. Therapeutic misconception. People with a new complete spinal cord injury are among the most decision-fragile populations in medicine. A consent process that says regenerative therapy but does not say there is no controlled evidence of efficacy is not adequate.
2. A repeat of the stem cell tourism pattern. Exosomes have inherited the marketing infrastructure built for unproven stem cell therapy — with an added advantage for the seller: a cell-free product sounds safer and rhetorically, though not legally, sidesteps some tissue regulation.
3. Cost and equity. Products are priced between thousands and tens of thousands of dollars, are not reimbursed, and are marketed to a population with high unemployment and high care costs.
4. Opportunity cost. The FDA notification makes this point explicitly: unlicensed products harm patients not only directly but also by displacing legitimate care. A chronic tetraplegic patient who spends their savings on intrathecal exosomes in 2026 is probably forgoing an approved, evidence-backed alternative that would help them.
5. Allogeneic donor consent — international standards for umbilical cord and placental tissue are variable.
6. Publication ethics. Reporting significant p values from roughly 15 uncorrected comparisons in a single-arm study of nine patients, in abstracts that are read far more than the limitations section, materially contributes to downstream harm even when the full text is rigorous.

Areas of consensus and of continuing debate

Clinical Recommendations and Patient Counselling

A patient counselling script

Direct Answers to Twelve Questions

A Future Research Roadmap

The design of the next decisive trial

Conclusion

Intrathecal exosome therapy in spinal cord injury occupies an unusual position: it is at once one of the better-conceived ideas in regenerative neurology and one of the more aggressively overhyped products of commercial medicine. Both statements are true, and the distance between them is the subject of this review.

The scientific rationale is real. Mesenchymal stromal cell benefit is largely paracrine; extracellular vesicles carry that paracrine signal; they cross biological barriers that cells cannot; they carry no tumorigenic or embolic risk; they can be stored, transported and re-administered. In rodent contusion models they reproducibly reduce inflammation and apoptosis, stabilise the blood–spinal cord barrier and improve locomotor scores. The first human study showed that a single intrathecal dose was tolerated over twelve months. An engineered product targeting the PTEN growth brake has obtained orphan drug status from two major authorities and is moving towards a placebo-controlled trial.

The evidential case, however, has not yet been made. That first human study enrolled nine patients without a control arm, in exactly the window of injury in which about one in five ASIA A patients recovers spontaneously; the sensory endpoints that reached significance were assessed unblinded, and no multiplicity correction was applied across fifteen comparisons. The preclinical literature motivating it shows measurable publication bias. The delivery assumption underlying the whole approach — that vesicles injected at L4/L5 reach the cord parenchyma at a thoracic lesion — has been demonstrated neither in humans nor in primates, and the best available primate data suggest meningeal rather than parenchymal distribution. There is no potency assay; which means that even if the product works, no regulator could license it. And the mechanisms exosomes actually engage are neuroprotective, which makes their use in chronic complete injury — the most heavily marketed population — a mechanistic contradiction over and above being an evidence problem.

Three things should follow. First, the field needs a potency assay and large animal biodistribution data far more urgently than it needs another Phase I. Second, the next decisive trial must be randomised and controlled; the Phase 2a component of ExoPTEN is the right decision point, and results preceding it should not be regarded as informative about efficacy. Third, and most urgently, clinicians should see this: the treatment is being sold to patients today, at high cost, with documented harms and without a licensed product anywhere in the world — and in 2026 there exists, for chronic incomplete tetraplegia, a regulator-cleared alternative that most of those patients have never heard of. The question is not whether exosomes are interesting. They are. The question is whether spinal cord injury research can show, in a field where patients are unusually fragile and commercial incentives point the other way, the discipline it has repeatedly failed to show. That question remains open.

Glossary

References

1. Akhlaghpasand, M., Tavanaei, R., Hosseinpoor, M., et al. (2024). Safety and potential effects of intrathecal injection of allogeneic human umbilical cord mesenchymal stem cell-derived exosomes in complete subacute spinal cord injury: A first-in-human, single-arm, open-label, phase I clinical trial. Stem Cell Research & Therapy, 15(1), 264. https://doi.org/10.1186/s13287-024-03868-0
2. Fehlings, M. G., Tetreault, L. A., Hachem, L., et al. (2024). An update of a clinical practice guideline for the management of patients with acute spinal cord injury: Recommendations on the role and timing of decompressive surgery. Global Spine Journal, 14(3_suppl), 174S–186S. https://doi.org/10.1177/21925682231181883
3. Tetreault, L. A., Kwon, B. K., Evaniew, N., Alvi, M. A., Skelly, A. C., & Fehlings, M. G. (2024). A clinical practice guideline on the timing of surgical decompression and hemodynamic management of acute spinal cord injury. Global Spine Journal, 14(3_suppl). https://doi.org/10.1177/21925682231183969
4. Badhiwala, J. H., Wilson, J. R., Witiw, C. D., et al. (2021). The influence of timing of surgical decompression for acute spinal cord injury: A pooled analysis of individual patient data. The Lancet Neurology, 20(2), 117–126. https://doi.org/10.1016/S1474-4422(20)30406-3
5. Bracken, M. B., Shepard, M. J., Holford, T. R., et al. (1997). Administration of methylprednisolone for 24 or 48 hours or tirilazad mesylate for 48 hours in the treatment of acute spinal cord injury (NASCIS III). JAMA, 277(20), 1597–1604.
6. Fehlings, M. G., Vaccaro, A., Wilson, J. R., et al. (2012). Early versus delayed decompression for traumatic cervical spinal cord injury (STASCIS). PLoS ONE, 7(2), e32037.
7. Global, regional, and national burden of spinal cord injury from 1990 to 2021 and projections for 2050: A systematic analysis for the Global Burden of Disease 2021 study. (2024). Ageing Research Reviews. https://doi.org/10.1016/j.arr.2024.102585
8. Karacan, İ., Koyuncu, H., Pekel, Ö., et al. (2000). Traumatic spinal cord injuries in Turkey: A nation-wide epidemiological study. Spinal Cord, 38(11), 697–701. https://doi.org/10.1038/sj.sc.3101064
9. Karamehmetoğlu, Ş., Ünal, Ş., Karacan, İ., et al. (1995). Traumatic spinal cord injuries in Istanbul, Turkey: An epidemiological study. Spinal Cord, 33(8), 469–471. https://doi.org/10.1038/sc.1995.102
10. Karamehmetoğlu, Ş. S., Nas, K., Karacan, İ., et al. (1997). Traumatic spinal cord injuries in southeast Turkey: An epidemiological study. Spinal Cord, 35(8), 531–533.
11. Aimetti, A. A., Kirshblum, S., Curt, A., et al. (2019). Natural history of neurological improvement following complete (AIS A) thoracic spinal cord injury across three registries. Spinal Cord. PMID 31182786
12. Spiess, M. R., Müller, R. M., Rupp, R., et al. (2009). Conversion in ASIA impairment scale during the first year after traumatic spinal cord injury. Journal of Neurotrauma, 26(11), 2027–2036.
13. Khorasanizadeh, M., Yousefifard, M., Eskian, M., et al. (2019). Neurological recovery following traumatic spinal cord injury: A systematic review and meta-analysis. Journal of Neurosurgery: Spine, 30(5), 683–699.
14. Ahuja, C. S., Wilson, J. R., Nori, S., et al. (2017). Traumatic spinal cord injury. Nature Reviews Disease Primers, 3, 17018.
15. Anderson, M. A., Burda, J. E., Ren, Y., et al. (2016). Astrocyte scar formation aids central nervous system axon regeneration. Nature, 532(7598), 195–200.
16. Bradbury, E. J., & Burnside, E. R. (2019). Moving beyond the glial scar for spinal cord repair. Nature Communications, 10(1), 3879. https://doi.org/10.1038/s41467-019-11707-7
17. Courtine, G., & Sofroniew, M. V. (2019). Spinal cord repair: Advances in biology and technology. Nature Medicine, 25(6), 898–908.
18. Liu, K., Lu, Y., Lee, J. K., et al. (2010). PTEN deletion enhances the regenerative ability of adult corticospinal neurons. Nature Neuroscience, 13(9), 1075–1081.
19. Park, K. K., Liu, K., Hu, Y., et al. (2008). Promoting axon regeneration in the adult CNS by modulation of the PTEN/mTOR pathway. Science, 322(5903), 963–966.
20. Welsh, J. A., Goberdhan, D. C. I., O'Driscoll, L., et al. (2024). Minimal information for studies of extracellular vesicles (MISEV2023). Journal of Extracellular Vesicles, 13(2), e12404. https://doi.org/10.1002/jev2.12404
21. Kalluri, R., & LeBleu, V. S. (2020). The biology, function, and biomedical applications of exosomes. Science, 367(6478), eaau6977. https://doi.org/10.1126/science.aau6977
22. Wiklander, O. P. B., Nordin, J. Z., O'Loughlin, A., et al. (2015). Extracellular vesicle in vivo biodistribution is determined by cell source, route of administration and targeting. Journal of Extracellular Vesicles, 4(1), 26316.
23. In vivo tracking of [89Zr]Zr-labeled engineered extracellular vesicles by PET reveals organ-specific biodistribution based upon the route of administration. (2022). Nuclear Medicine and Biology. https://doi.org/10.1016/j.nucmedbio.2022.06.004
24. Wang, L., Pei, S., Han, L., et al. (2018). Mesenchymal stem cell-derived exosomes reduce A1 astrocytes via downregulation of phosphorylated NFκB P65 subunit in spinal cord injury. Cellular Physiology and Biochemistry, 50(4), 1535–1559.
25. Xin, W., Qiang, S., Jianing, D., et al. (2021). Human bone marrow mesenchymal stem cell-derived exosomes attenuate blood–spinal cord barrier disruption via the TIMP2/MMP pathway after acute spinal cord injury. Molecular Neurobiology, 58, 6490–6504.
26. Fan, L., Dong, J., He, X., Zhang, C., & Zhang, T. (2021). Bone marrow mesenchymal stem cells-derived exosomes reduce apoptosis and inflammatory response during spinal cord injury by inhibiting the TLR4/MyD88/NF-κB signaling pathway. Human & Experimental Toxicology, 40(10), 1612–1623.
27. Yi, H., & Wang, Y. (2021). A meta-analysis of exosome in the treatment of spinal cord injury. Open Medicine, 16(1), 1043–1060. https://doi.org/10.1515/med-2021-0304
28. Stem cell-derived exosome treatment for acute spinal cord injury: A systematic review and meta-analysis based on preclinical evidence. (2025). Frontiers in Neurology, 16, 1447414. https://doi.org/10.3389/fneur.2025.1447414
29. Efficacy of miRNA-modified mesenchymal stem cell extracellular vesicles in spinal cord injury: A systematic review and network meta-analysis. (2022). Frontiers in Bioengineering and Biotechnology. https://doi.org/10.3389/fbioe.2021.819651
30. Bydon, M., Qu, W., Moinuddin, F. M., et al. (2024). Intrathecal delivery of adipose-derived mesenchymal stem cells in traumatic spinal cord injury: Phase I trial (CELLTOP). Nature Communications, 15. https://doi.org/10.1038/s41467-024-46259-y
31. Vaquero, J., Zurita, M., Rico, M. A., et al. (2018). Intrathecal administration of autologous mesenchymal stromal cells for spinal cord injury: Safety and efficacy of the 100/3 guideline. Cytotherapy, 20(6), 806–819.
32. El-Kheir, W. A., Gabr, H., Awad, M. R., et al. (2014). Autologous bone marrow-derived cell therapy combined with physical therapy induces functional improvement in chronic spinal cord injury patients. Cell Transplantation, 23(6), 729–745.
33. Weidner, N., Abel, R., Maier, D., et al. (2025). Safety and efficacy of intrathecal antibodies to Nogo-A in patients with acute cervical spinal cord injury: A randomised, double-blind, multicentre, placebo-controlled, phase 2b trial (NISCI). The Lancet Neurology, 24(1), 42–53. https://doi.org/10.1016/S1474-4422(24)00447-2
34. Moritz, C., Field-Fote, E. C., Tefertiller, C., et al. (2024). Non-invasive spinal cord electrical stimulation for arm and hand function in chronic tetraplegia: A safety and efficacy trial (Up-LIFT). Nature Medicine. ClinicalTrials.gov NCT04697472
35. Lorach, H., Galvez, A., Spagnolo, V., et al. (2023). Walking naturally after spinal cord injury using a brain–spine interface. Nature, 618(7963), 126–133.
36. ONWARD Medical N.V. (2024, 19 December). ARC-EX System receives FDA de novo authorization.
37. ONWARD Medical N.V. (2025, 17 November). ONWARD Medical receives FDA 510(k) clearance to expand the ARC-EX System indication for home use.
38. U.S. Food and Drug Administration. (2019, 6 December). Public safety notification on exosome products. https://www.fda.gov/vaccines-blood-biologics/safety-availability-biologics/public-safety-notification-exosome-products
39. U.S. Food and Drug Administration. (2026, March). Patient and consumer warning about potential serious risks of harm following use of unapproved products from human cells or tissues. https://www.fda.gov/vaccines-blood-biologics/safety-availability-biologics/patient-and-consumer-warning-about-potential-serious-risks-harm-following-use-unapproved-products
40. Rahnama, M., Heidari, M., Poursalehi, Z., & Golchin, A. (2024). Global trends of exosomes application in clinical trials: A scoping review. Stem Cell Reviews and Reports, 20(8), 2165–2193. https://doi.org/10.1007/s12015-024-10791-7
41. Wei, W., Wu, X., Wang, L., et al. (2026). The global clinical landscape of human-derived extracellular vesicles: An analysis based on ClinicalTrials.gov (2010–2025). Clinical and Experimental Medicine, 26, 200. https://doi.org/10.1007/s10238-026-02124-4
42. Turkish Medicines and Medical Devices Agency. Guideline on advanced therapy medicinal products. https://www.titck.gov.tr
43. Betül-Ziya Eren Genome and Stem Cell Centre, Erciyes University. (2026, July). Manufacturing Site Authorisation Certificate for the production of umbilical cord-derived mesenchymal stem cell exosomes [institutional announcement].
44. ExoPTEN: Allogeneic exosome therapy for spinal cord injury with strong therapeutic potential and clinical promise. (2025). Cytotherapy. https://doi.org/10.1016/j.jcyt.2025.03.132
45. NurExone Biologic Inc. (2026, 1 June). First quarter 2026 financial results and corporate update [press release].
46. Anderson, K. D. (2004). Targeting recovery: Priorities of the spinal cord-injured population. Journal of Neurotrauma, 21(10), 1371–1383.
47. Simpson, L. A., Eng, J. J., Hsieh, J. T., & Wolfe, D. L. (2012). The health and life priorities of individuals with spinal cord injury: A systematic review. Journal of Neurotrauma, 29(8), 1548–1555.
48. Itzkovich, M., Gelernter, I., Biering-Sorensen, F., et al. (2007). The Spinal Cord Independence Measure (SCIM) version III: Reliability and validity in a multi-center international study. Disability and Rehabilitation, 29(24), 1926–1933.
49. Krogh, K., Christensen, P., Sabroe, S., & Laurberg, S. (2006). Neurogenic bowel dysfunction score. Spinal Cord, 44(10), 625–631.