A baby’s umbilical cord becomes medical waste along with the placenta. Across its cut surface run two arteries and one vein, and a transparent jelly fills the space around them. This is Wharton’s jelly (WJ). It is a tissue built on the same idea as the cushioning in a parcel, and packed densely inside it are mesenchymal stem cells (MSCs). They are the youngest cells there are, from the very day of birth.
What we are reading today is a review published in Brain Sciences by a team at the National Scientific Medical Center in Astana, Kazakhstan. Wharton’s jelly mesenchymal stem cells (WJ-MSCs) and their secreted products are better suited to post-stroke regenerative medicine than the “adult” MSCs of bone marrow or fat — that is the claim, and the conclusion section calls them an “innovative breakthrough in regenerative medicine.”
First, the character of this paper. It is a review that carried out not a single new experiment. The paper itself states explicitly that it “neither generated nor analyzed new data,” and the “Methods” field of the abstract is a single sentence: to “discuss the current state of MSCs.” A search string, inclusion and exclusion criteria, and a PRISMA (the international standard that sets out the reporting procedure for systematic reviews) flow diagram are not reported. How the 285 references were chosen cannot be verified by the reader. And here is the heaviest fact of all. Among those 285 references, there is not a single clinical trial in which cells or exosomes explicitly identified as “Wharton’s jelly-derived” were given to stroke patients. And this despite the subtitle being “Current Role of WJ-MSCs.”
Let me settle the words before we read. The cord’s formal name is the umbilical cord, and “umbilical cord-derived MSC” (UC-MSC) lumps three different things together: (1) cells from the jelly (WJ-MSC), (2) cells from umbilical cord blood, and (3) cells from minced whole umbilical cord. The review’s argument begins at the point where this distinction goes slack. And the secretome is the totality of soluble factors, conditioned medium is the cultured medium itself, an extracellular vesicle (EV) is a membrane sac, and an exosome is one of those that came from the endosomal pathway. These four are not synonyms.
Publication Details
- Paper title: Stem Cells in Post-Stroke Regenerative Therapy: Current Role of Wharton’s Jelly Mesenchymal Stem Cells in the Orchestrum
- Paper type: Labeled “Review” in the journal. It is a narrative review in which no search string, no selection criteria, and no PRISMA flow are reported
- Authors: Anastassiya Ganina (first author and corresponding author) and others, eight in all
- Affiliation: JSC National Scientific Medical Center in Astana, Kazakhstan. All authors are from a single institution
- Journal: Brain Sciences (MDPI), 2026, Volume 16, Issue 8, Article No. 775
- DOI / Link: 10.3390/brainsci16080775
- Peer review and publication dates: Received May 29, 2026; accepted July 22; published July 23 (54 days to acceptance, 1 day to publication)
- Open access: Yes (CC BY)
- Impact Factor: 3.4 (Journal Citation Reports released June 2026, data year 2025). This is the publisher’s stated figure, and it is not written in the review
- About the journal: Indexed in SCIE (Science Citation Index Expanded) and Scopus, but it is not a MEDLINE-indexed journal, so the standard index terms of MeSH (Medical Subject Headings) are not assigned to it (confirmed in the catalogue of the US National Library of Medicine; information from outside the review)
- Funding: Program-targeted funding from the Ministry of Science and Higher Education of Kazakhstan (Grant No. BR28713159). Its title is “Innovative approaches toward the creation of cell products”
- Conflicts of interest: A single line, “there is no conflict of interest.” A review written on funding from a national program to create cell products is promoting that very class of product
About corresponding author Anastassiya Ganina (what follows comes from outside the review, confirmed against public information): A researcher in the cell technology and transplantation department of her own institution. Her specialty is the bioengineering that carries stem cells, and cells for cancer immunotherapy, into clinical application. Ten years of practical experience and nine patents are listed on the institutional site. The official profile and the department page disagree with each other, so her title cannot be pinned down. Experience leading a cell therapy trial genuinely does exist — she is the Kazakhstan-side principal investigator of an international collaborative trial using autologous adipose-derived regenerative cells for chronic ischemic heart failure (ClinicalTrials.gov NCT06705023). But that is the heart, not the brain. No prior stroke research of hers turns up in PubMed, OpenAlex, or ORCID. The one self-citation among the 285 references is her own case report in which allogeneic WJ-MSCs were given intradermally for chronic postoperative wounds (Ganina A, et al., Diseases 2026;14(1):27) — she is in the position of providing, at her own institution, the very modality she promotes. Her own group’s paper on EVs and lyophilization (Journal of Clinical Medicine of Kazakhstan 2025;22(6):117–125) is a review that puts “lyophilization” in its keywords and asserts its effect in the conclusion, and yet it is cited nowhere among the 285 references.
What Wasn’t Understood Before?
Acute stroke treatment has grown stronger over these 20 years. Intravenous thrombolysis, when given within 3 hours of onset, gives an odds ratio of 1.75 (95% CI 1.35–2.27) for a modified Rankin Scale (mRS; 0 is no symptoms, 6 is death) of 0 or 1 at 3 to 6 months. Past 4.5 hours the significant difference disappears (Emberson J, et al., Lancet 2014;384(9958):1929–1935). Thrombectomy is stronger still: across 5 trials and 1287 patients the adjusted common odds ratio was 2.49 (1.76–3.53), p<0.0001, and the number needed to treat (how many patients must be treated for one person to benefit) was 2.6 (Goyal M, et al., Lancet 2016;387(10029):1723–1731). “Stroke cannot be treated” is no longer correct.
The problem is that few people make it in time. Eligibility for thrombectomy covers roughly 1 in 10 stroke admissions (McMeekin P, et al., Eur Stroke J 2017;2(4):319–326). Above all, recanalization therapy does not bring back neurons that have already died. The number of people alive today who have experienced a stroke is about 94 million worldwide (Feigin VL, et al., Int J Stroke 2025;20(2):132–144).
The first hope — “replace the dead neurons” — stumbled in animal experiments. In a rat model of transient cerebral ischemia, MSCs given intravenously did not reach the brain, while those given arterially did reach it but raised the risk of brain lesions, and functional recovery did not improve (Argibay B, et al., Sci Rep 2017;7:40758). When allogeneic bone marrow MSCs were given to 38 rats through the external carotid artery, cerebral blood flow fell in a dose-dependent manner, emboli increased, the lesion expanded, and sensorimotor deficits worsened (Cui LL, et al., Stem Cell Res Ther 2015;6(1):11). As far as rodents show, the safe route does not deliver, and the route that delivers is dangerous.
👦 Student: If cells given intravenously don’t reach the brain, isn’t there no point in giving cells at all?
🧬 Dr. Exotaro: This is the idea that has changed most in 20 years. Stop sending the carpenters; send the site foreman’s written instructions instead. The cells don’t have to take up residence. Scatter the instruction sheets called brain-derived neurotrophic factor (BDNF) and glial cell line-derived neurotrophic factor (GDNF), and the surviving neurons, the vascular endothelium, and the microglia (the immune cells resident in the brain) start repairing things themselves. This review, too, summarizes it as: the regenerative effect is “achieved primarily through paracrine effects,” not through transdifferentiation. If so, the next step is this — don’t send the carpenters at all, send only the instruction sheets. That is the review’s answer.
What Did This Paper Find?
”Adult MSCs” and “Young MSCs” — The Opposition the Review Sets Up
To harvest bone marrow is to harvest cells worn in by as many years as that person has lived. The cells of Wharton’s jelly are from the day of birth, and they are not swayed by the patient’s age or comorbidities — the flip side of what Table 2 of the review writes, that “bone marrow and adipose MSCs are strongly influenced by age and comorbidities.” But this is the claim of a table cell with not a single reference attached, and variability on the donor side does not vanish. The review itself writes that the composition of the secretome and of EVs “varies with donor status, cell source, and culture conditions.” Youth has its price too. Bone marrow and fat can both be taken from the patient, yet Wharton’s jelly always ends up being someone else’s cells — the review, too, writes that “WJ-MSCs are always an allogeneic material.” The strength and the price come out of the same fact.
Table 2 lines up WJ-MSCs, bone marrow MSCs (BM-MSCs), and adipose tissue MSCs (AD-MSCs) along 7 axes, assigning to BM-MSCs “lowest proliferative capacity, longest doubling time (the time it takes for the cell number to double), replicative senescence at P5–P7 of passage (the operation of moving cells to a new dish to expand them),” and to WJ-MSCs “highest proliferative capacity, shortest doubling time” and “highest tolerogenicity.” But when you trace them through the main text, the basis for this “highest” and “shortest” is almost entirely comparisons inside a culture dish.
From the Cells to What the Cells Put Out
The story of youth turns here into the question of whether to give cells at all. As we saw above, in rodents cells given intravenously did not reach the brain, while the arterial route that did reach it plugged the fine vessels and widened the lesion. The review, too, writes that the capacity of living cells to cross the blood-brain barrier is “extremely limited” — it does not write that it is zero. If the substance of the effect is paracrine, there is no need for a courier large enough to cause a blockage. It is on this logic that the review pushes the secretome and exosomes forward.
Give no cells, and turn only what they put out into the product — the advantages the review lists for this move to cell-free are scalable production and precise dose setting through lyophilization and ultrafiltration, minimization of immunogenicity, vascular embolism, and tumorigenic risk because no cells are given, and becoming an “off-the-shelf” product, ready to use the moment it is needed — all of them claims on paper, not advantages verified in stroke patients. The review writes the price as well — standardized collection protocols and dedicated facilities, and the fact that only by giving exosomes many times over does one get an effect comparable to giving cells.
👦 Student: The “vascular embolism” you just mentioned — does it mean that the cells given to treat a cerebral infarction end up creating a new cerebral infarction?
🧬 Dr. Exotaro: That’s what it means. But what has been seen clearly is an experiment with intra-arterial administration in rodents, and how much it happens in humans is unknown. Against the brain’s finest vessels, cells are something like gravel. Exosomes, on the other hand, are 30–150 nm in diameter according to Table 1 of the review. That is a millionth of a millimeter, so seen from inside a vessel they are not gravel but mist. This is the most easily grasped rationale for “giving only the secreted products” — and it is at the stage of rationale, not an advantage verified in humans. And the problem of “does it reach?” is not dispelled by this. The review introduces, in animals, EVs gathering at the infarct and moving into the damaged tissue. But the number for how much actually reaches the brain itself after intravenous injection in humans appears nowhere.
The Active Ingredient of the “Secreted Product” Is Never Settled
This is why I separated the four words at the start. The review defines the secretome as “consisting of numerous soluble factors,” and writes that the therapeutic effect is “mediated by the secretome and exosomes, that is, by the secretion of soluble factors and extracellular vesicles.” The definitions themselves are not contradictory. What fails to line up comes after. Throughout, they are bundled with a slash as “secretome/exosomes,” and results from unfractionated conditioned medium get restated as “the secretome was superior.” Which fraction is to be called the active ingredient is never settled.
👦 Student: It looks like a fussy point about wording — is it really that important?
🧬 Dr. Exotaro: To formulate a product is to decide what you define as the active ingredient. Is it the soluble protein, the vesicle, or the microRNA (miRNA) inside the vesicle? Until that is settled, you cannot write specifications, a potency assay, or lot release criteria. The review correctly writes that “batch-to-batch variation is a challenge.” But it never defines the very thing whose variation is to be measured.
Table 1 defines exosomes by their biogenetic origin (the endosomal pathway), but MISEV2023, the international guideline for EV research, does not recommend giving a collection of isolated vesicles a name based on “where they were born” — once they have been isolated, the place of birth cannot be verified (Welsh JA, et al., J Extracell Vesicles 2024;13(2):e12404). This review never once mentions MISEV2023, nor the International Society for Extracellular Vesicles (ISEV) that issued it. The claim that “exosomes are small, so they cross the blood-brain barrier (BBB)” is stacked on top of a definition of origin that has not been demonstrated.
The Preclinical Work Is Moving. But the Human Column Stays Empty
The animal data are not thin. When exosomes derived from human WJ-MSCs were given in a rat model of ischemic stroke, TUNEL-positive (dead) neurons decreased (Chiu YS, et al., Cell Transplant 2024;33:9636897241296366). But rewriting creeps in. The review writes that “WJ-MSCs” loaded with miR-9-5p reduced infarct area in “an experimental rat model of ischemic stroke,” yet the title of the cited source is a study using “human umbilical cord MSCs” in “hypoxic-ischemic brain injury in neonatal rats” (Li B, et al., Brain Behav 2026;6:e71282). Whether the cells were WJ-derived cannot be judged from the title, but the rewriting of the model can be seen from the title alone. Neonatal hypoxia-ischemia differs from the adult condition in both its pathophysiology and its therapeutic target.
And in humans? The only WJ-MSC trial described in detail is a Phase I trial in traumatic brain injury (TBI) (Kabatas S, et al., World J Stem Cells 2024;16(6):641–655). It is not stroke. The authors themselves write that trials using umbilical cord-derived cells (UC-MSC / WJ-MSC) in stroke are “only 20%” (Fan Y, et al., Stem Cells Transl Med 2026;15:szaf082), and that of 50 trials in neurodegenerative disease, those using WJ-MSCs number “only 2” (Isik S, et al., CNS Neurosci Ther 2025;31:e70577). And yet the first sentence of the “Highlights” this journal displays reads: the regenerative capacity of MSC therapy, particularly that using WJ-MSCs, “is supported by numerous preclinical studies and clinical trials.” The main text, meanwhile: although recovery generally improves in animals, “clinical trials of BM-MSCs and AD-MSCs often report minimal or zero effect.” They write by name that WJ-MSCs are “supported by clinical trials,” and the contents of those clinical trials are the results of cell sources other than WJ-MSCs. Most readers read only the Highlights.
👦 Student: So has nobody yet had a stroke cured with Wharton’s jelly?
🧬 Dr. Exotaro: Precisely speaking, the answer is “within this review there is not a single stroke clinical trial explicitly stated to be Wharton’s jelly-derived.” I cannot assert that the number is zero. Outside the review there is a report from Poland in which 3×10⁷ WJ-MSCs were given repeatedly by the intrathecal route (into the fluid surrounding the spinal cord, from the back) to 6 patients with chronic-stage stroke, recording improvement in the 4 who received three or more doses (Milczarek O, et al., Cell Transplant 2023;32:9636897231195145). But it is a non-randomized case series of 6 versus 6, set alongside 6 patients on standard care alone, and moreover it is not among the review’s 285 references.
How Will the Future Change? (The Path to the Clinic)
This is a review. In stroke, WJ-MSCs have come only as far as preclinical work (rodents) — what is described in detail in humans is a Phase I trial in traumatic brain injury. A search of the registries turns up one stroke trial of WJ-MSC-derived EVs in the world — STEVIA (ClinicalTrials.gov NCT06995625, an open-label, single-arm, dose-escalation trial in South Korea). n=18, three cohorts at 4.8×10¹⁰ / 9.6×10¹⁰ / 19.2×10¹⁰ particles, starting June 22, 2026 — one month before this review was accepted, and the review does not mention it. For the cells themselves, not a single WJ-MSC stroke trial is registered (this is the result of a registry search). The EV trials are all Phase I or Phase I/II, and not one of them has published results. Cell therapy, on the other hand, has climbed as far as Phase II/III and fallen.
👦 Student: If it can be freeze-dried and kept on a shelf, then like instant ramen, anyone can use the same thing, right?
🧬 Dr. Exotaro: The metaphor fits the convenience, but it doesn’t fit the uniformity. The autologous cell therapies that have been tried in stroke are fresh pasta delivered to your door. Once the patient is decided, you harvest bone marrow and culture it over several weeks. If the acute-phase contest runs 18 to 36 hours from onset, no amount of compressing makes those several weeks arrive in time. But there is one place where it breaks. A factory can make instant ramen identical every time, but MSC-derived products cannot be made that way yet. It is “we want it to be so,” not “it already is.”
What ought to be decided before starting is left blank as well. The dose of EVs is never once stated — not as particle number, not as protein amount, not as cell equivalents (the EVs harvested from how many cells’ worth). The only dose printed anywhere is “up to 1.5×10⁶ cells per kg of body weight.” To hold up “precise dose setting can be done” as an advantage while showing no units is not coherent. The timing of administration is the same. The review introduces the “therapeutic window” not as the time elapsed from onset but in the sense of “what stage that MSC is at within the culture at the moment of administration,” and uses it in the sense of time in the thrombolysis context. Timing on the patient’s side is not specified, and Table 2 says “acute and subacute phase” — an unconstrained span running from the day of onset to 6 months. Neither the “selected patients” of the Highlights, nor the stroke severity scale (National Institutes of Health Stroke Scale, NIHSS; the higher, the more severe), nor infarct volume is defined. On hemorrhagic stroke, which accounts for three in ten of all strokes, the cell therapy chapter only notes that “there are trials enrolling ischemic/non-ischemic patients,” and does not examine its specific efficacy and safety.
How to Critically Read This Research (Limitations and Paths to Improve It)
Before criticism, the virtues. The structure itself is built as a contrast: because Chapter 3 puts “everything other than WJ-MSCs” first, the reader is able to ask, “couldn’t it be something other than WJ?” Treating cell-free products as an independent modality and writing all the way through manufacturing, logistics, and cost; holding the end point to “an adjunctive therapy for selected patients” rather than “cure”; writing four negative clinical trials into the main text; and not dropping citations that work against their own position are good judgments too.
The list of limitations the authors themselves acknowledge is, if anything, honest. (1) Results are contradictory, conspicuously so for long-term effect. (2) The main focus of existing trials is safety and tolerability. (3) Dose, cell type, route of administration, and assessment criteria vary widely between studies. (4) WJ-MSCs are always an allogeneic material, and clinical use is rare. The three most damaging to themselves are written down as well. (5) Intravenous administration is as safe as the other routes, but its long-term results are inferior to them (a meta-analysis of 17 studies). (6) Repeated intravenous administration of EVs may, unlike live cells, accelerate the clearance of the EVs themselves — the single most inconvenient fact for their own premise that “only by giving many doses does one match cells.” It is not a contradiction, but it calls for a design for the number of doses. (7) Which assessment scale is chosen strongly determines the measured result — they have written half of the primary endpoint problem themselves.
On the other hand, there are a number of important problems the paper does not touch on.
The outline of the negative trials is blurred. The review never once uses the term “primary endpoint” in its entire text. The trial names, too, appear in the reference list (STARTING-2, AMASCIS, TREASURE) but not in the main text. The one that “enrolled more than 200 patients at 18 to 36 hours from onset and found no difference on any measure used” is TREASURE (Houkin K, et al., JAMA Neurol 2024;81(2):154–162). Beyond that, MASTERS, MASTERS-2, PISCES-2, and ACTIsSIMA are not cited at all. While the review writes that “intracranial administration gives the best recovery,” it does not touch on the fact that the only large trial to have tested that route with a sham surgery control (SB623, 163 patients) missed its primary endpoint. The route of administration, too, is never settled. Three syntheses made under different conditions are lined up — for umbilical cord MSCs, intravenous administration gives better NIHSS and FIM than intrathecal administration (one clinical study); intracranial administration gives the best recovery but is the most dangerous (a systematic review of 21 reports); and intravenous administration is as safe as the other non-systemic routes but inferior in long-term results (a meta-analysis of 17 studies). The latter two are compatible with each other, but they do not mesh with the first. The review merely acknowledges the inconsistency as a general matter, and does not do the work of setting the three against one another. Until the route is settled, “an effective adjunctive therapy for selected patients” cannot be defined.
The same negative trial reappears a few sentences later as positive evidence. The review first writes that “intravenous administration of preconditioned autologous MSCs did not improve mRS at 3 months” ([122] = STARTING-2, Chung JW, et al., Neurology 2021;96(7):e1012–e1023). Yet further on in the same paragraph, as “Similarly, intravenous administration of preconditioned autologous MSCs affords protection of the corticospinal tract and better recovery,” the same [122] is cited bundled together with a neuroimaging sub-analysis. Nowhere in the main text is there any explanation that “these two are the same trial.” The paragraph closes: “the main limitations of these studies were small sample sizes and incomplete blinding.” That reads as “the sample was too small to detect it,” but what the original reported was that “no difference emerged on the pre-specified primary endpoint.” The reason a single primary endpoint is fixed in advance is that if you are allowed to pick the best-looking of the measures taken afterwards, you can manufacture an “it worked” out of chance alone. In this reduction to generic limitations, the fact of the miss is diluted, and the secondary analysis is left standing like a conclusion.
What the reference title shows to be a clinical experience report of 5 patients, and the negative result of a Phase II/III trial in more than 200, are lined up in the same paragraph as coordinate declarative sentences (Ercelen N, et al., Front Med 2023;10:1051831). Neither the sample size nor the design of the former appears in the main text. With a risk-of-bias assessment, they could not have been placed like this. The most rigorous synthesis is missing too. The Cochrane review (Boncoraglio GB, et al., Cochrane Database Syst Rev 2019;5(5):CD007231) concluded that stem cell transplantation was associated with a reduction in neurological impairment but was not associated with improvement in functional outcome, and graded the certainty (the grade of how far one may believe that conclusion) as low to very low. Without this, there is no way to convey why “the meta-analyses are positive” and “the Phase III trials missed” can both hold at once. Another omission is a synthesis that answers head-on the review’s own question, “which cell source is the best.” A network meta-analysis comparing 7 kinds of stem cell therapy across 18 randomized trials and 1,006 patients in total (BMC Neurology 2025) appears nowhere among the 285 references. Higher-level evidence exists, and yet the review answers with culture-dish data and a table with no sources.
There are several places where a citation does not match the claim in the main text. The main text writes that “labeled BM-MSCs migrated to the injured neurons of stroke rats and expressed NeuN,” citing [164], but [164] is a paper on exosomes protecting cardiomyocytes from acute myocardial infarction (Huang L, et al., Cell Cycle 2020;19:339–353). Conversely, [162], which the main text cites for myocardial infarction, is a brain paper dealing with blood-brain barrier disruption and hemorrhagic transformation (Yang J, et al., Mol Neurobiol 2026;63:351). But it is not a simple swap of numbers either. Elsewhere [162] does match the main text, and the reader has no way of knowing which reference ought to have been cited. The notation of cell source goes slack as well. The review writes “when WJ-MSCs are preconditioned with hypoxia” and cites two papers — rat distal middle cerebral artery occlusion (Abuzan M, et al., Transl Stroke Res 2025;16:817–830) and mouse cerebral ischemia (Gregorius J, et al., Basic Res Cardiol 2021;116:40). Both cited titles say “mesenchymal stromal cells,” and whether they are WJ-derived cannot be confirmed from the review. Unless the harvest site is checked in the original papers, they cannot be counted as evidence for WJ. Together with the miR-9-5p case above, this is a question of how far to trust the remaining 282 references that have not been verified.
The central argument never gets out of the culture dish. The comparisons cited to show that WJ is superior to the others are, when traced, almost all inside a culture dish (proliferation, 3D scaffolds, co-culture comparisons). And not a single head-to-head comparison in a stroke model animal, setting WJ-MSCs against other cell sources side by side within the same experiment, is cited. Some of the remaining animal data are precarious too. The review cites a study in a rat cerebral ischemia-reperfusion model and summarizes the reduction in infarct area and the improvement in memory and learning as its effects, but that study gave the WJ-MSCs before the occlusion (Abd El Motteleb DM, et al., J Cell Biochem 2018;119:9790–9800) — the review states this explicitly itself and does not hide it. And although it does not hide it, it adds not one word of note that this is preventive pretreatment and cannot serve as a basis for a treatment given after onset. The starting point is precarious as well. “In an in vitro comparison, UC-MSCs survived longest in culture and BM-MSCs were the poorest” — this is the first basis for “the younger the cell, the higher its proliferative capacity,” yet the cited source is a comparison of bone marrow, umbilical cord blood, and fat, not Wharton’s jelly (Kern S, et al., Stem Cells 2006;24:1294–1301). The three things distinguished at the outset dissolve at the very starting point of the argument. The review writes that amniotic epithelial stem cells surpass WJ-MSCs in immune tolerance and antioxidant activity, and also that cells from the junction of the umbilical cord and the placenta express pluripotency markers at higher levels than WJ-derived cells. If “the younger and less differentiated, the better” is taken literally, the optimal answer is not WJ-MSCs but amniotic epithelial cells or junction cells. The central argument cannot derive the authors’ conclusion.
The tables and the main text contradict each other. The main text writes that the doubling time is “higher than that of BM-MSCs” (= slower proliferation), while Table 2 says “shortest.” They are exact opposites. Neither Table 1 nor Table 2 carries a single reference, so “highest,” “shortest,” and “a dramatic reduction in infarct volume” are all unsourced. For surface antigens, CD24, CD105, CD108, CD73, CD90 are listed, but there is no panel of the negative markers used to identify MSCs (the marks that must not be there). What there is is a single line, “they do not express markers specific to fibroblasts,” from which the review proceeds to “a fraction with little contamination by other cells is obtained.” Purity can only be claimed once both positive and negative markers are shown. CD24 is not a standard MSC marker in the first place, and ruling out fibroblasts alone cannot exclude contamination by hematopoietic or endothelial cells. The “low risk of graft-versus-host disease (GvHD)” that Table 2 lists as an advantage of WJ also does not say low compared with what. GvHD occurs in treatments such as bone marrow transplantation, where the donor’s immune cells are given along with everything else and those immune cells attack the patient’s body. Neither BM-MSCs nor AD-MSCs, the comparators, are given along with immune cells. What actually becomes a problem in MSC infusion runs the other way — rejection, in which the patient’s immunity eliminates the MSCs, and sensitization, in which repeated administration generates antibodies against the donor’s leukocyte type (HLA). Those are not in the table. In the clinical trial column for WJ-MSCs, Table 2 also writes that “preference for them as an allogeneic stroke treatment is increasing.” Preference can increase even where use is rare, so this does not collide with the main text’s “WJ-derived use is rare” and “2 out of 50 trials.” But for this “is increasing” there is neither a figure showing the trend nor a source. The “highest tolerogenicity” claim can be read as being about the three cell sources, but there is no source for that ranking.
The assertions in the conclusion section carry no basis. Section 6 contains not a single reference, and there it is written in the present tense that “lyophilized cellular contents promote recovery of neurovascular coupling (the coordination between neurons and blood vessels) and improve neuromotor function.” But the citations attached to lyophilization in the main text number three — a topical gel of WJ-MSC secretome, GMP (Good Manufacturing Practice)-compliant lyophilized production of BM-MSC secretome, and the development of lyophilized “off-the-shelf” cell-free products (Driscoll J, et al., Pharmaceutics 2022;14:849) — and not one of them showed efficacy in a stroke model. And not a single p-value, effect size, or confidence interval is printed anywhere. Although neither a search string nor a bias assessment is shown, the Author Contributions assign “methodology,” “validation,” “formal analysis,” and “data curation,” and what work those refer to cannot be read off from the paper.
The safety issues specific to stroke are left blank as well. In the acute phase of ischemic stroke, antithrombotic drugs are used as standard, and some patients are only just out of thrombolysis or thrombectomy. Not one line is written about a protocol for how to time the combination. Neither surveillance for undiagnosed malignancy in an elderly population nor endotoxin and residual serum components in conditioned medium-derived products is raised. The review writes that MSCs may accumulate in the tumor microenvironment and promote tumorigenesis, and holds that going cell-free avoids this, but EVs, too, carry angiogenic factors and growth-promoting miRNAs.
So how could the quality have been raised? In order of what works at zero cost. (1) Write the actual search into the “Methods” of the abstract. (2) Make the Highlights and the conclusion agree with the main text. (3) Put a citation number and a figure into every cell of Table 1 and Table 2 — the cells that cannot be filled become the list of adjectives that should be cut. (4) Add a column to the clinical trial table for “did it meet its primary endpoint?” — the most honest figure is one in which the WJ-MSC rows all come out blank. (5) Tag every citation with its cell source (WJ / whole umbilical cord / cord blood / MSCs in general). (6) Align the terminology with the MSC definition (Dominici M, et al., Cytotherapy 2006;8(4):315–317) of the International Society for Cell & Gene Therapy (ISCT) and with MISEV2023 — standardize MSC as mesenchymal stromal cell; move from “pluripotent” MSCs to “multipotent” MSCs (expression of NANOG, Oct3/4, and Sox2 is not proof of pluripotency; iPS cells and embryonic stem cells stay as they are); add the negative markers (CD45, CD34, CD14/CD11b, CD79a/CD19, HLA-DR) to the surface antigens; and call vesicles whose biogenesis has not been demonstrated small EVs. The apparent strength drops, but that is the level of the evidence. (7) Decide on a potency assay (a test that measures the strength of the active ingredient) and measure batch-to-batch variation. (8) Apply RoB 2 to the clinical trials (ROBINS-I for non-randomized studies) and SYRCLE risk-of-bias assessment to the animal experiments, and grade the certainty with GRADE for each of mRS, NIHSS, Barthel, and adverse events — then the real picture, “the meta-analyses are positive but the certainty is low to very low, and the large trials missed their primary endpoints,” can be written in one line. (9) Rewrite “selected patients” into inclusion criteria (the reason for restricting to ischemic stroke, the NIHSS range, age, infarct volume, days since onset, and 90-day mRS as the primary endpoint). (10) Declare at the outset whether hemorrhagic stroke is being covered or excluded (and if excluded, the reason why promoting angiogenesis could be harmful in a hemorrhagic condition). (11) Fix the mismatched citations, and write what each role in the Author Contributions referred to.
And one decisive move as well. Compare EVs derived from WJ-MSCs, BM-MSCs, and AD-MSCs in the same stroke model animal, with the dose matched by particle number, with passage number, culture conditions, and isolation method unified, with randomization, an a priori sample size calculation, and blinding of the assessors in place, and with the protocol pre-registered (and reported in accordance with SYRCLE and ARRIVE 2.0). Put in a group of living WJ-MSCs as well, and the question of “cells or cell-free” is answered at the same time. And do it in both young and old animals — since Table 2 claims that “BM-MSCs from elderly donors are less effective,” the recipient’s aging must be tested too. Stroke patients are elderly, after all. This does not weaken the authors’ conclusion. Because “for the other cell sources, large trials that failed to meet their primary endpoints are lined up; the clinical efficacy of WJ-MSCs in stroke is not yet established, but preclinically they are promising” is, precisely in being verifiable, the stronger claim of the two.
Dr. Exotaro’s Perspective
I myself have continued to work on applying MSC-derived small extracellular vesicles (sEVs) to the treatment of spinal cord injury (SCI). The shift in this review’s center of gravity is not somebody else’s problem to me.
In a study we reported earlier, in a rodent model of severe spinal cord injury, the MSCs given intravenously never reached the injury site themselves, yet the sEVs those MSCs released inside the body migrated to the injury site and associated specifically with the M2 macrophages (macrophages that have switched their character to the side that calms inflammation) residing there (Nakazaki, M., et al., Journal of Extracellular Vesicles 2021;10(11):e12137). The two things this review writes — that the efficacy of cells and of cell-free preparations is essentially equivalent (the basis is a single scoping review, and whether its quality was assessed cannot be told from the review), and that only by giving exosomes many times over does one obtain an effect comparable to giving cells — are both exactly what I myself saw in rodents (the sEVs were given in three divided doses).
But that is what I saw in the rodent spinal cord. In this series we have read human trials using cells in stroke three times now. The double-blind, placebo-controlled trial that gave allogeneic (donor-derived) cells intravenously in the acute phase came out at p=0.83 on the composite assessment at 90 days (the figures in this paragraph do not appear in this review). The single-arm trial of 18 patients that stereotactically transplanted cells into the chronic-phase brain (placing them at one point inside the brain by fixed coordinates) “met” its primary endpoint, yet in the 163-patient trial with a sham surgery control the result was p = 0.6743. A single-arm trial of 23 patients that put neural stem cells into the putamen (deep in the cerebrum) did not reach it either. In the first of those trials the immune biomarkers clearly moved, yet at 90 days nothing had changed. That a mechanism moves and that a person gets better are separate questions.
What this repetition of negative results teaches, I think, is not that the mechanism is wrong but that the problem lies in the kind of cell, the dose, the timing, the route, and the way patients are selected. The review itself writes that “large-scale multicenter randomized trials, accompanied by rigorous standardization of cell type, dose, timing, and route of administration, are required.” It writes the correct prescription itself, and supplies an answer to none of the items.
There are places where optimism outruns the evidence. This review takes as its premise that “exosomes work because they cross the BBB,” but there is a competing account holding that the amount of systemically administered EVs that reaches the brain parenchyma is small, and that the effect is better explained by reprogramming the immunity of the periphery (Doeppner TR, et al., Stem Cells Transl Med 2015;4(10):1131–1143). Does it work because it reaches the brain, or does it work by quieting systemic inflammation without ever reaching it? The review does not pose this question.
There are still reasons to place the bet. That cells given intravenously are caught in the lung and spleen, that they do not engraft (settle in), immunogenicity, vascular embolism — with secreted products, to exactly the extent that no cells are given, all of these become smaller in principle — and this is not a superiority verified in human stroke. And “becomes smaller” is not “disappears.” Even in the biodistribution study the review cites, accumulation of EVs in the spleen, lung, and liver is merely said to be lower than that of living cells; it is not written as zero. The review itself also writes that WJ-MSC-derived exosomes contain proteins involved in complement and coagulation activation at high concentrations. With risk, only the kind of risk is swapped out. Even so, the difference is not small. Part of the reason cell therapy could not produce an answer may have been a problem not of biology but of logistics and uniformity. But as for the head-to-head comparison I demanded of the authors, it cannot be said that this field — my own rodent work included — has ever satisfied it.
To readers who are patients or their families, let me write this plainly. As of September 2026, there is no drug or regenerative medical product established by high-quality controlled trials to reliably restore the sequelae of stroke themselves. And no results from a human controlled trial showing that Wharton’s jelly-derived exosomes work in stroke have been published either. The Japanese Society for Regenerative Medicine, too, has issued a caution that cell therapy given outside insurance coverage “has not necessarily passed review for safety and therapeutic efficacy” (May 2022). A caution about stem cells applies, just as it stands, to exosomes as well.
In its conclusion the review uses the phrase “innovative breakthrough.” I do not use it. Not because I do not believe in this field. The cell-free products of WJ-MSCs, too, must one day receive their answer under the same severity as cell therapy. So that we can say “it worked” when that time comes, count the particles, define the active ingredient, and design in advance the control that would make the effect disappear. This road is not yet closed. But it is not open either.
Note: this article was written with the assistance of AI.