Human clinical evidence on stem cell–based therapies for ischemic stroke: An umbrella review with emphasis on mesenchymal stem cells

Abstract

Objective

To synthesize human clinical evidence from systematic reviews and meta-analyses on stem cell–based therapies for ischemic stroke, with an emphasis on mesenchymal stem cell–based interventions, and to appraise their efficacy, safety, methodological quality, overlap, and certainty of evidence.

Methods

This Open Science Framework–registered umbrella review followed Preferred Reporting Items for Systematic Reviews and Meta-Analyses 2020 and Joanna Briggs Institute guidance. PubMed/MEDLINE, Scopus, and Web of Science were searched through 1 November 2025. Eligible studies were systematic reviews or meta-analyses of randomized or nonrandomized human clinical studies evaluating mesenchymal stem cells or related stem cell–based interventions for ischemic stroke. Two reviewers independently performed study selection, data extraction, AMSTAR 2 (A MeaSurement Tool to Assess systematic Reviews 2) appraisal, and GRADE (Grading of Recommendations Assessment, Development and Evaluation)-based certainty assessment. Because of overlap among reviews and heterogeneity in cell products, administration routes, timing, and outcomes, the findings were synthesized narratively rather than through a de novo meta-analysis.

Results

Twenty-six reviews were included. The evidence suggested possible improvements in neurological impairment, disability, activities of daily living, and motor recovery; however, effect estimates varied according to cell type, administration route, timing after stroke, and follow-up duration. Intravenous administration was the most frequently studied and appeared feasible, with acceptable short-term safety reporting, whereas intra-arterial and intracerebral routes showed greater procedure-related concerns. Short-term adverse events were not consistently increased. However, the evidence remains insufficient regarding rare or delayed risks, including tumorigenesis, ectopic tissue formation, embolic events, and alloimmunogenicity. The certainty of evidence was low to moderate for most efficacy outcomes, moderate for short-term adverse events, and very low to low for long-term or rare safety outcomes.

Conclusions

Stem cell–based therapies, particularly mesenchymal stem cell–based approaches, may improve selected neurological and functional outcomes after ischemic stroke. However, overlapping reviews, heterogeneous interventions, small early-phase trials, and limited long-term follow-up require cautious interpretation. Adequately powered, standardized, multicenter trials with long-term safety surveillance are needed before routine clinical implementation.

Introduction

Ischemic stroke remains a leading cause of death and long-term disability worldwide despite advances in reperfusion therapy and organized stroke systems of care. ¹ Even with timely thrombolysis and thrombectomy, many survivors experience persistent neurological deficits, limited independence in activities of daily living, and inadequate recovery with conventional rehabilitation alone, underscoring the need for neurorepair strategies that extend beyond acute recanalization. ² Mesenchymal stem cell (MSC)–based therapies have emerged as a biologically plausible approach to promote recovery through multimodal mechanisms, including immunomodulation, proangiogenic and neurotrophic signaling, and support of endogenous neuroplasticity and white matter repair. ³ Contemporary narrative and quantitative syntheses describe MSCs as capable of attenuating post-ischemic inflammation, reducing apoptosis, and enhancing angiogenesis and synaptic remodeling, thereby improving functional outcomes across several phases of stroke recovery. ⁴ Preclinical studies also indicate that non-MSC stem/progenitor populations, such as epidermal neural crest stem cells, may influence infarct volume and neurorepair-related gene expression pathways after experimental ischemic stroke. Therefore, findings from MSC-based interventions should not be generalized to all stem cell products without product-specific evidence. ⁵

Since 2021, several systematic reviews and meta-analyses have evaluated human clinical evidence for stem cell–based therapies, particularly MSC-based interventions, in ischemic stroke. ⁶ These reviews suggest possible improvements in neurological impairment, disability, activities of daily living, and motor recovery. ⁷ However, interpretation remains limited by small early-phase trials, heterogeneity in cell products and delivery protocols, variable follow-up durations, and overlap among the primary studies included in different reviews.

Evidence regarding administration route and treatment timing also continues to evolve. Intravenous administration is the most frequently investigated and appears more feasible for clinical use, whereas intra-arterial and intracerebral approaches may show greater effects in selected contexts but involve greater procedure-related complexity and risk. Similarly, subacute and chronic treatment windows may be associated with greater functional gains than hyperacute administration in some analyses, but these findings require confirmation in adequately powered comparative trials. ⁸ Beyond stroke, MSC-based therapies have been evaluated across a broad clinical trial landscape involving autoimmune, inflammatory, and degenerative diseases, highlighting both the translational appeal of MSCs and the continuing need for standardized trial design, product characterization, and outcome reporting. ⁹

Notwithstanding these encouraging signals, the literature remains heterogeneous with respect to cell source (bone marrow, umbilical cord, and adipose tissue), manufacturing and potency assays, administration route and timing, and follow-up duration, complicating cross-study comparisons and evidence grading. Recent systematic review/meta-analyses emphasize that future progress depends on harmonized chemistry, manufacturing, and controls (CMC); head-to-head evaluations of leading strategies; and longer-term safety surveillance and maintain a strong emphasis on methodological rigor, including International Prospective Register of Systematic Reviews (PROSPERO)/Open Science Framework (OSF) registration, duplicate review processes, Cochrane-compatible RoB, and GRADE (Grading of Recommendations Assessment, Development and Evaluation). ¹⁰

Against this background, we conducted a prospectively registered umbrella review of systematic reviews and meta-analyses of human ischemic stroke treated with MSC-based interventions. Our goals were as follows: (a) summarize efficacy across neurological impairment, disability/activities of daily living (ADL), and motor recovery; (b) integrate safety outcomes, including adverse events, tumorigenesis, embolic events, and mortality; (c) appraise methodological quality using AMSTAR 2 (A MeaSurement Tool to Assess systematic Reviews 2) and certainty of evidence using GRADE; and (d) identify route- and timing-specific considerations and evidence gaps to guide the next generation of Phase III and implementation-focused trials.

Methods

Protocol and registration

This umbrella review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines^11,12 and the Joanna Briggs Institute (JBI) recommendations for umbrella reviews. The protocol outlining the review objectives, eligibility criteria, search strategy, and planned analytical approach was prospectively registered on the OSF to ensure transparency and methodological rigor. The registration record can be accessed on the OSF using the following DOI: 10.17605/OSF.IO/ZY9HP.

Any subsequent modifications to the protocol were documented and updated in the OSF record. The protocol served as the guiding framework for the entire review process, including database searching, data extraction, quality assessment, and evidence synthesis.

Eligibility criteria

We included systematic reviews and/or meta-analyses that evaluated the efficacy and/or safety of MSCs or related stem cell–based therapies in human ischemic stroke populations. Eligible reviews synthesized randomized controlled trials (RCTs), nonrandomized clinical studies, or mixed clinical designs and reported at least one clinical, functional, imaging, or safety outcome. MSC-focused reviews were the primary target of this umbrella review. Reviews that included broader stem cell products were eligible only when human ischemic stroke outcomes were reported and when MSC- or stem cell–based clinical data could be extracted separately or interpreted within clearly defined intervention categories.

For interpretability, interventions were classified into the following categories: (a) MSC-based products, including bone marrow–, umbilical cord–, and adipose-derived MSCs; (b) bone marrow mononuclear cells; (c) progenitor cell products, including endothelial progenitor cells, peripheral blood stem cells, multipotent adult progenitor cells, cluster of differentiation 34 (CD34)–positive cells, and aldehyde dehydrogenase (ALDH)–positive cells; (d) neural or genetically modified cell products; and (e) mixed or combination approaches. Findings from non-MSC or mixed-cell reviews were treated as indirect evidence and interpreted separately.

We excluded narrative reviews, scoping reviews, preclinical-only reviews, animal-only studies, nonstroke indications, duplicate publications, commentaries, editorials, conference abstracts, and reviews lacking sufficient methodological transparency. We restricted inclusion to peer-reviewed full-text articles published in English. This restriction was applied for feasibility and interpretability but was considered a potential source of language bias.

Information sources and search strategy

A comprehensive literature search was conducted to identify all relevant systematic reviews and meta-analyses evaluating MSC therapy in human ischemic stroke. Three major electronic databases were systematically searched: PubMed, Scopus, and the Web of Science (WOS). The search covered all available studies published through 1 November 2025, with English-language restrictions applied during the initial screening phase. The search strategy combined controlled vocabulary terms and free-text keywords related to the intervention and condition (Table 1).

Table 1.

Search strategy used for PubMed, Scopus, and Web of Science.

Database	Search query	Date of search	Notes
PubMed (via MEDLINE)	((“mesenchymal stem cells”[MeSH Terms] OR “mesenchymal stem cell”[Title/Abstract] OR “mesenchymal stromal cell”[Title/Abstract] OR “MSC”[Title/Abstract]) AND (“ischemic stroke”[MeSH Terms] OR “stroke, ischemic”[Title/Abstract] OR “cerebral infarction”[Title/Abstract] OR “brain ischemia”[Title/Abstract]) AND (“meta-analysis”[Publication Type] OR “systematic review”[Publication Type] OR “meta analysis”[Title/Abstract] OR “systematic review”[Title/Abstract]))	1 November 2025	Filters: Humans; English; No time restriction.
Scopus (Elsevier)	TITLE-ABS-KEY((“mesenchymal stem cell*” OR “mesenchymal stromal cell*” OR MSC) AND (“ischemic stroke” OR “cerebral infarction” OR “brain ischemia”) AND (“meta-analysis” OR “systematic review”))	1 November 2025	Searched in Article Title, Abstract, and Keywords; no year restriction; English.
Web of Science (Clarivate, Core Collection)	TS = ((“mesenchymal stem cell*” OR “mesenchymal stromal cell*” OR MSC) AND (“ischemic stroke” OR “cerebral infarction” OR “brain ischemia”) AND (“meta-analysis” OR “systematic review”))	1 November 2025	All years; all document types; English only.

MESH: medical subject heading.

The reference lists of included reviews and relevant primary studies were also manually screened to identify additional eligible publications. All retrieved records were imported into EndNote 20 for duplicate removal and organized according to the PRISMA 2020 guidelines.

Study selection

All records identified through database searching were imported into EndNote 20 for reference management and duplicate removal. The study selection process was performed in accordance with the PRISMA 2020 framework and carried out in two sequential stages: title and abstract screening followed by full-text evaluation.

Two reviewers independently screened all titles and abstracts to determine preliminary eligibility. Full-text articles of potentially relevant studies were then assessed independently by the same reviewers using the predefined inclusion and exclusion criteria. Any discrepancies or disagreements in study selection were resolved through discussion and consensus, and when necessary, consultation with a third reviewer to ensure methodological transparency and minimize selection bias.

Data extraction

Data extraction was conducted independently by two reviewers using a standardized data collection form designed for this umbrella review. For each included systematic review or meta-analysis, the following key information was extracted: first author and publication year; study design (systematic review or meta-analysis); type of included studies (RCTs or clinical trials); population characteristics (adult patients with ischemic stroke); intervention details (MSC type and source); route of administration (intravenous, intra-arterial, intracerebral, or intrathecal); and comparators (placebo or standard therapy).

Primary and secondary outcomes were also extracted, including measures of neurological function (National Institutes of Health Stroke Scale (NIHSS), modified Rankin Scale (mRS), and Barthel Index (BI)); motor and functional recovery (Fugl-Meyer assessment (FMA) and functional independence measure (FIM)); imaging or biomarker outcomes (infarct volume and angiogenic and neurogenic markers); and safety outcomes (adverse events, serious adverse events, and mortality). When available, pooled effect estimates (mean difference, standardized mean difference, odds ratio, or risk ratio) were recorded together with the corresponding 95% confidence intervals and heterogeneity statistics (I²). All extracted data were cross-verified for accuracy between reviewers, and discrepancies were resolved through discussion and consensus to ensure consistency and reliability in the final synthesis.

Methodological quality assessment

The methodological quality of all included systematic reviews and meta-analyses was evaluated using AMSTAR 2. ¹³ . This tool assesses 16 key methodological domains related to protocol registration, adequacy of the literature search, risk of bias assessment, statistical methods, and reporting transparency. Each review was independently appraised by two reviewers, and any disagreements were resolved through discussion and consensus. Overall AMSTAR 2 ratings were categorized as high, moderate-high, moderate, or low, reflecting confidence in the methodological rigor of each review.

The certainty of evidence was assessed using the GRADE framework at the umbrella-review level. When included reviews reported GRADE ratings, these were extracted as supporting information; however, the final certainty judgments were determined independently by two reviewers for this umbrella review. The assessment considered the methodological quality of the included reviews, risk of bias in the underlying clinical studies (when reported), inconsistency, indirectness, imprecision, publication bias, and overlap of primary studies across reviews. Because many primary studies were small early-phase trials and several reviews included overlapping primary trials, the certainty of evidence was interpreted conservatively. Disagreements were resolved through discussion or consultation with a third reviewer.

Data synthesis

Findings were synthesized narratively according to outcome domain, including neurological impairment, disability, ADL, motor recovery, imaging or repair-related outcomes, adverse events, serious adverse events, mortality, and rare or delayed safety outcomes. For each included review, we extracted the reported effect estimates, confidence intervals, heterogeneity statistics, subgroup findings, and certainty assessments when available.

No de novo meta-analysis, quantitative reanalysis of primary studies, recalculation of pooled effect estimates, funnel plot analysis, Egger test, trim-and-fill analysis, or robust variance estimation was performed in this umbrella review. This decision was made because the included reviews shared overlapping primary trials, used different effect measures and follow-up time points, and evaluated heterogeneous cell products, administration routes, and treatment windows. Therefore, the interpretation was based on the direction and consistency of the reported effects, methodological quality, GRADE certainty, and overlap among reviews.

Management of overlap across reviews

To assess overlap among the included reviews, we mapped the primary clinical studies included in each systematic review or meta-analysis and calculated the corrected covered area (CCA). Primary-study identifiers were normalized to reduce double counting of duplicate reports or follow-up publications from the same trial. The CCA was calculated as follows:

$$\mathrm{CCA}=(\mathrm{N}-\mathrm{r})/(\mathrm{r}\times \mathrm{c}-\mathrm{r})\times 100$$

where N is the total number of primary-study occurrences across reviews, r is the number of unique primary studies, and c is the number of included reviews. CCA values were interpreted as indicating slight (0%–5%), moderate (>5%–10%), high (>10%–15%), or very high (>15%) overlap. The overlap assessment informed the GRADE certainty judgments and was one reason that a de novo pooled meta-analysis across reviews was not performed. Detailed overlap mapping is provided in Supplementary Table S1.

Results

Search results

A total of 152 records were identified through database searching, including 54 from WOS, 89 from Scopus, and 9 from PubMed/MEDLINE. Manual screening of reference lists did not identify any additional eligible records. After the removal of 49 duplicate records, 103 records remained for title and abstract screening.

Following title and abstract screening, 76 records were excluded because they were not systematic reviews or meta-analyses, did not evaluate ischemic stroke, did not include human clinical evidence, or did not assess eligible stem cell–based interventions. Therefore, 27 full-text reports were retrieved and assessed for eligibility. Of these, one report was excluded because the required data were unavailable or insufficient. Finally, 26 systematic reviews and meta-analyses met the inclusion criteria and were included in the umbrella review and qualitative synthesis (Figure 1).

Figure 1.

PRISMA 2020 flow diagram of study identification and selection. The diagram summarizes the database search, duplicate removal, title and abstract screening, full-text eligibility assessment, and the final inclusion of systematic reviews and meta-analyses evaluating stem cell–based therapies for ischemic stroke.

PRISMA: Preferred Reporting Items for Systematic Reviews and Meta-Analyses.

Characteristics of included reviews

This umbrella review included 26 systematic reviews and meta-analyses published between 2011 and 2025 (Table 2).^{6–8,10,14–35} Most reviews synthesized randomized or controlled human clinical studies, whereas a small number also included mixed clinical and preclinical evidence; for these reviews, human ischemic stroke data were considered separately where possible.^14–16 Two reviews used network meta-analysis to compare cell sources or administration routes.^17,18

Table 2.

Characteristics of included meta-analyses on mesenchymal stem cell therapy for ischemic stroke in human clinical studies.

Author, year (ref)	Region	Study type	n (primary)	N (MSC/CTRL)	MSC source(s)	Routes	Comparator	Outcomes	Model	Key findings (direction and key effect sizes)	Heterogeneity (I2)	Pub. bias
Alrasheed et al. 6	Saudi Arabia	RCTs (MA)	18	1026 (538/488)	BM-MSC, AD-MSC, UC-MSC, EPC, OEC, MAPC	IV (main), IA, IN, SC	Placebo/Std care	NIHSS, mRS, BI, FMA, Infarct, AEs	REM	mRS ↓ (MD −0.56); BI ↑ (MD +12); FMA ↑ (MD +18.2); NIHSS ≈; SAEs ≈ (RR 1.03).	NIHSS 91%; mRS 17%; BI 69%; FMA 80%; Infarct 67%	Funnel plots neg
Boncoraglio et al. 19	Italy	RCTs (MA)	7	401 (188/213)	BM-MNC, BM-MSC, MAPC, PBSC	IV, IA, IC, LSAS	Placebo/Std care	NIHSS, mRS, BI, safety	REM	NIHSS ↓ (MD −1.49); mRS ≈ (MD −0.42, NS); BI ↑ (NS); mortality ≈; overall safe, modest neuro benefit.	NIHSS 66%; mRS 72%; BI 78%	Not done (<10)
Cao and Li 20	China	RCTs + NRCTs (MA)	5	228 (104/124)	Autologous BM-MSC (2), BM-MNC (3)	IV (4), IA (1)	Std care	NIHSS, mRS, BI/mBI, AEs	FE/RE	NIHSS ↓ (MD −1.85); BI ↑ (MD +7.44); mRS ≤2 ≈; safety ≈.	NIHSS 24%; BI 46%; mRS 57%	Not done (<10)
Chang et al. 17	China	RCTs (NMA + MA)	19	1055 (total)	BM-MNC, BMSC, PC, PBSC, UBMSC	IV, IA, SAH, stereotaxic	CRT	NIHSS, mRS, MBI, FMA, death/AEs	NMA (SUCRA)+REM	Rankings: UBMSC best for NIHSS; BM-MNC best for mRS/MBI/FMA. Mortality ≈; SAH ≤2 wk +>12 months FU favored.	NIHSS 84%; mRS 50%; MBI 98%; FMA 17%	Low
Chumnanvej and Chumnanvej 21	Thailand	RCTs + comp. (MA)	9	469 (231/238)	BM-MNC (±BM-MSC)	IV (4), IA (4), IC (1)	Std care	NIHSS, BI, mRS≤3, death, AEs	FE/RE	NIHSS ≈ at 3–6 months; BI ↑ at 6 months (SMD +1.17); death/AEs ≈; IA route improved NIHSS.	NIHSS 95%; BI 93%; mRS 12%	NR
Fauzi et al. 8	Indonesia	RCTs + NRCTs (MA)	21	817 (406/411)	BM-MSC (65%), PBSC, UC, AD, NSC	IC (5), IA (3), IV (13)	Std therapy	NIHSS, mRS, BI, AEs/SAEs	RE	IC route: greatest NIHSS/mRS gains at 6 months; IV: BI ↑; AEs overall 26%; SAEs highest with IC; risk ratios for AEs/SAEs ≈ vs. control.	AEs 72%; SAEs 0%	Bias tables adequate
Hassanein et al. 22	Egypt/DE/IRQ	RCTs (MA)	4	97 (≈49/48)	BM-MSC, AD-MSC	IV only	Placebo/Std care	NIHSS, mRS, BI, mortality, recurrence	FE	NIHSS ≈; mRS ↓ (MD −0.95); BI ↑ (MD +21.4); mortality/recurrence ≈.	NIHSS/mRS/BI 0%; mort 0%; recur 36%	Not done (<10)
Huang et al. 23	China	RCTs (MA)	9	316 (159/147)	BM-MSC (8), AD-MSC (1)	IV/IA (mostly IV)	Placebo/Std care	NIHSS, mRS, BI, AEs	RE	NIHSS ↓ (SMD −0.99); mRS ≈; BI ≈; AEs ≈ (RR 0.68).	NIHSS 73%; mRS 63%; BI 73%; AEs 56%	Egger neg
Kumar et al. 7	India	RCTs (MA)	8	459 (217/242)	BM-MSC, BM-MNC, PBSC, MAPC	IV (4), SAH (3), IA (1)	Placebo/Std care	NIHSS, mRS, BI, mortality	FE/RE	Chronic phase: NIHSS ↓ (SMD −1.57), mRS ↓ (SMD −1.07); acute/subacute ≈; meta-regression: SAH route and chronic phase better.	NIHSS acute 67%; chronic 0%; mRS chronic 72%; BI 93%	Not done
Kurniawan et al. 24	Indonesia	Non-RCTs (SR)	4	69 (approximately 69/0; 1 RCT)	BM-MSC, AD-MSC (allo + auto)	IV (2), IC stereotaxic (2)	Placebo (1)/single-arm	NIHSS, FMA, BI, mRS, AEs	Descriptive	Chronic stroke: consistent motor and ADL gains (FMA ↑ ∼19 at 12–24 months); AEs mild; no deaths/tumors; small/heterogeneous.	NA	Not done
Kurniawan et al. 25	Indonesia	RCTs (SR)	8	563 (155/408)	MSC/MNC (mostly autologous)	IV, IA, IT	Std therapy	NIHSS, mRS, BI, imaging, AEs	Descriptive	Acute IS: MNC improved NIHSS within 1 month; MSC delayed gains (3–6 months); imaging infarct expansion ↓; safety ≈.	—	Not done
Kvistad et al. 26	Norway	Controlled (MA)	9 IS studies (5 RCTs)	171 (101/70)	Auto BM-MSC; allo UC-MSC	IV (main), IT, intralesional	Placebo/Std care	mRS (mean and 0–2), NIHSS, AEs/SAEs	RE	Safety: 3 SAEs (surgical route–related); IV largely safe. Efficacy: mRS ≈ (MD −0.11), mRS0–2 ≈.	mRS 0–2 42%; mean mRS 51%	Not done
Lalu et al. 14	Canada	Mixed (MA)	10 clinical + 76 preclin	Clinical 339 (187/152)	Mixed (BM 84%, UC 9%, AD 3%…)	IV, IC, IA, IN	Placebo/Std rehab	Mortality, NIHSS, mRS, BI, motor tasks	RE	Clinical: mortality ↓ (Peto OR 0.43); fever risk ↑ (OR 6.88); efficacy overall mixed/no clear benefit. Preclinical: robust functional gains.	Preclin 88%–91%; Clinical NP	Egger: small-study bias (adhesive removal)
Moñivas Gallego and Zurita Castillo 27	Spain	Trials registry (SR)	14 trials (planned/ongoing)	approximately 680 planned	UC (7), BM (4), AD (2), NS (1)	IV (78%), IT, IA, intraparenchymal	Placebo/std arms or open-label	Safety, feasibility, NIHSS/mRS/BI	Descriptive	Acute-phase trials: safety and trend to neuro improvement; few significant functional gains yet; two possibly related SAEs resolved.	—	—
Mulia et al. 28	Taiwan/Indonesia	Clinical (MA)	17	959 (intervention 669; CTRL 922)	BM-MNCs, MSCs, EPCs, PBSCs	IV 59%, non-IV 41% (IA/IC/LP)	Placebo/std	NIHSS, BI, mRS, mortality, AEs	RE	NIHSS ↓ (6-month SMD −0.42); BI ↑ (SMD +0.41); mRS ↓ (6-months SMD −0.39); mortality ↓ (OR 0.48); AEs ≈; non-IV > IV for long-term function.	NIHSS 58%; mRS <75%; AEs <75%	Low
Ouyang et al. 15	China	Mixed (MA)	11 clinical + 11 preclin	Clinical ∼500	MSCs, MNCs, NSCs, PBSCs	IV (main), IA, IC, SAH, stereotaxic	Placebo/std	NIHSS, BI, mRS, FMA, safety	RE	Clinical: NIHSS ↓ (MD −2.57), BI ↑ (+7.93), mRS ↓ (−0.53), FMA ↑ (+5.5); mild fever ↑; best with early IV, UC-MSC, ≤107 cells.	NIHSS 51%; BI 59%; mRS 0%; FMA 15%	NR
Permana et al. 10	Indonesia	RCTs + NRCTs (MA)	19 (4 pooled)	800 (379/421)	BM-MSC, BM-MNC, PBSC, NSC, UC-MSC	IV (most), IA, IC, SAH	Placebo/std	NIHSS, mRS, BI (6–24 months), AEs	RE	6–12 months benefit: NIHSS ↓ (−1.48 at 6 months), mRS ↓ (−0.53 at 12 months); BI ≈; signal plateaus by 24 months; safety good.	NIHSS 83%; mRS 80%; BI 84%	ROBINS-I/RoB2 acceptable
Raghuwanshi et al. 29	India	RCTs (SR)	6	190 (115/75)	Auto MSC (2), G-CSF (3)	IV (MSC), SC (G-CSF)	Placebo/std rehab	Motor NIHSS, FMA, MI, RMA, EMS, fMRI, mRS	Narrative	Mixed motor effects across trials; overall safe; heterogeneity precluded pooling.	—	JBI only
Shen et al. 30	China	RCTs + NRCTs (MA)	30 (15/15)	1217 (624/593)	BM-MSC, AD-MSC, UC-MSC, PBSC, EPC	IV (main), IA, IC, IN	Placebo/std	mRS, NIHSS, BI, FMA, mortality, AEs	RE	mRS ↓ (MD −0.26); NIHSS ↓ (MD −1.69); mortality ↓ (RR 0.44); BI/FMA ≈ overall; IV effective; window 2 w–3 m; autologous > allogeneic; AEs mild.	NIHSS 27%; mRS 30%; BI 24%	Funnel plots neg
Unsworth et al. 31	Australia	Mixed clin. ≤90 d (MA)	28	1808 (675/598/535 pre-post)	BM-MSC, BM-MNC, UC-MSC, NSC, G-CSF, CD34+	IV, IC, IT, IA, SC	Std rehab/placebo	NIHSS, BI, FMA, FIM, MRI, SAEs	RE	Largest gains with IT NSCs (g 2.85) and IV + SC BM-MSC + G-CSF (g 2.56); IV BM-MSC g 0.88; safety acceptable; age <60 and higher dose linked to larger effects.	Post 92.6%; Pre-post 96.3%	Fail-safe N > n
Unsworth et al. 32	Australia	Mixed clin. > 90 d (MA)	23	485 (206/279 pre-post)	NSC, BM-MSC, BM-MNC, UC-MSC, PBSC, NT2/D1, SB623, G-CSF	IC, IT, IV, SC, combos	Std care/rehab	NIHSS, mRS, BI/FMS, SAEs	RE	IT NSCs and IV BM-MNCs show largest functional gains; IC route SAEs mainly surgical; higher dose (≥108) better.	Post 70.3%; Pre-post 85.1%	Fail-safe N > n
Wang et al. 18	China	RCTs (NMA/MA)	13	704 (total)	BM-MSC, UC-MSC, MAPC, EPC, CD34+, MNC, ALDH+	IV, IA, IC, IT	Std care	mRS, BI, NIHSS, Mortality, SAEs	Bayesian RE (MCMC)	mRS ↓, BI ↑, NIHSS ↓, mortality ↓, SAEs ≈. SUCRA: IA MNC/ALDH+ best for ADL; IC CD34+ best neuro; IV MSC/EPC best survival/SAEs.	mRS 86%; BI 81%; NIHSS 79%; SAEs 47%	No bias
Wang et al. 33	China	Controlled (MA)	7	288 (142/146)	Autologous BM-MSC; PBSC	IV (4), IA (1), IC (1), stereotaxic (1)	Std/placebo	Mortality, NIHSS, ESS, BI, mRS, FM, ESSMS, Ashworth	FE/RE	ESS and ESSMS ↑; NIHSS/BI/mRS ≈; mortality ≈; small-dose NIHSS signal.	NIHSS 62%; mRS 68%; BI 39%	Egger/Begg neg
Xue et al. 34	China	RCTs + NRCTs (MA)	23	1279 (625/654)	BM-MSC (18), UC-MSC (3), UCB-MSC (2)	IV (14), SAH (5), IT (3), IA (1)	Std rehab/medical	NIHSS, BI, FMA, FIM, AEs	RE	NIHSS ↓ (3 months −2.65); BI ↑ (6 months +10.1); FMA ↑ (6 months +10.8); FIM ↑ (6 months +16.6); safety good.	NIHSS 85%; BI 81%; FMA 98%; FIM 88%	Funnel plots neg
Zhang and Wang 35	China	RCTs (MA)	13	692 (376/336)	BM-MSC	IV, IA, SAH, IT	Std medical/rehab	NIHSS, FMA, BI, FIM, mRS, AEs	FE/RE	NIHSS ↓ (3 months −2.88); FMA ↑ at 1/3/6 months; FIM ↑ at 1/3/6 months; BI ≈ at 3–6 months; mRS ≈ ≤6 months; AEs ≈.	NIHSS <50%; FMA/FIM/BI >50%	FIM bias possible
Zheng et al. 16	China/USA	Mixed (SR)	8 clin + 78 preclin	Clin 89; Preclin ∼900	BM-MSC (major), UC-MSC (2), others	IV, IA, IC, IT, IN	Std/sham	Clin: NIHSS, FMA, mRS, BI; Preclin: mNSS, infarct, motor	Qualitative	Clin: trends to benefit; AEs mild; Preclin: consistent robust benefit. Gaps in reporting and trial design noted.	—	—

AD: adipose-derived (cell/source); AD-MSC: adipose-derived mesenchymal stem cell; ADL: activities of daily living; AE: adverse event; AEs/SAEs: adverse events/serious adverse events; ALDH⁺: aldehyde dehydrogenase–positive (cell fraction); Bayesian RE (MCMC): Bayesian random-effects (Markov Chain Monte Carlo); BI: Barthel Index; BM-MNC: bone marrow mononuclear cells; BM-MSC/BMSC: bone marrow mesenchymal stem cells; comp.: comparative (controlled nonrandomized studies); CRT: conventional rehabilitation therapy; CTRL: control (group); DE: Germany (country code in row: Egypt/DE/IRQ); EMS: European Motor Scale (or equivalent motor subscore; see source study); EPC: endothelial progenitor cell; ESS: European Stroke Scale; ESSMS: European Stroke Scale Motor Subscale; FE: fixed-effects (meta-analysis model); FE/RE: fixed-effects/random-effects (both used); FIM: functional independence measure; FM: Fugl-Meyer (total); FMA: Fugl-Meyer assessment; FMS: Fugl-Meyer Score; fMRI: functional magnetic resonance imaging; FU: follow-up; IA: intra-arterial; IC: intracerebral (intracranial/intraparenchymal); IN: intranasal; IRQ: Iraq (country code in row: Egypt/DE/IRQ); IS: ischemic stroke; IT: intrathecal; IV: intravenous; JBI: Joanna Briggs Institute (risk of bias/critical appraisal); LSAS: lumbar subarachnoid space; MA: meta-analysis; MAPC: multipotent adult progenitor cells; MBI: Modified Barthel Index; MCMC: Markov Chain Monte Carlo; MI: Motricity Index; mNSS: modified neurological severity score; mRS: modified Rankin Scale; MSC: mesenchymal stem cell; NMA: network meta-analysis; NR: not reported; NRCTs: nonrandomized controlled trials; NS: not specified (cell/source); NSC: neural stem cell; OR: odds ratio; PBSC: peripheral blood stem cells; PC: progenitor cells; preclin: preclinical (animal); RE: random-effects (meta-analysis model); REM: random-effects model; RoB2: Cochrane risk of bias tool 2 (for RCTs); ROBINS-I: risk of bias in nonrandomized studies of interventions; RCTs: randomized controlled trials; RR: risk ratio; SAH: subarachnoid (route of administration); SB623: genetically modified BM-MSC product (SB623 line); SC: subcutaneous (route); SR: systematic review; Std/Std care/Std rehab: standard care/standard rehabilitation; SUCRA: Surface Under the Cumulative Ranking Curve (NMA ranking metric); UBMSC: umbilical cord blood mesenchymal stem cells; UC: umbilical cord (source); UC-MSC: umbilical cord mesenchymal stem cells; UCB-MSC: umbilical cord blood mesenchymal stem cells; wk: week(s); ref: reference; G-CSF: granulocyte colony-stimulating factor; NT2/D1: NTERA-2 clone D1 human teratocarcinoma-derived neural precursor cell line; OEC: olfactory ensheathing cell; MD: mean difference; SMD: standardized mean difference; NP: not provided; LP: lumbar puncture.

The symbol ≈ indicates no significant difference/comparable.

The clinical evidence covered the acute, subacute, and chronic phases of ischemic stroke. The most frequently evaluated products were bone marrow–derived MSCs and bone marrow mononuclear cells, followed by umbilical cord–derived MSCs, adipose-derived MSCs, endothelial progenitor cells, peripheral blood stem cells, multipotent adult progenitor cells, neural/progenitor cell products, and genetically modified cell products. Administration routes included intravenous, intra-arterial, intracerebral or stereotaxic, intrathecal or subarachnoid, intranasal, and subcutaneous approaches.

The most commonly reported outcomes were neurological impairment, disability, ADL, motor recovery, imaging or repair-related outcomes, adverse events, serious adverse events, and mortality. Sample sizes, follow-up durations, intervention characteristics, and outcome definitions varied substantially across reviews. Because many reviews included overlapping primary trials, cumulative participant counts across reviews were not interpreted as representing unique patients. Review-specific characteristics and reported estimates are presented in Table 2.

Methodological quality

Methodological quality was assessed using AMSTAR 2 (Figure 2). Eighteen reviews were rated as high, three as moderate-high, and five as moderate. No review was rated as low or critically low. Common methodological strengths included comprehensive literature searches, duplicate study selection and data extraction, risk of bias assessment, and adherence to established systematic review reporting standards.

Figure 2.

Methodological quality of the included systematic reviews and meta-analyses based on AMSTAR 2 ratings. Bubble sizes represent the relative risk score (1–4), and distinct labels or patterns indicate the quality categories (moderate, moderate-high, and high).

AMSTAR 2: A MeaSurement Tool to Assess systematic Reviews 2.

Common limitations included incomplete reporting of funding sources for the primary studies, limited discussion of how the risk of bias in the primary studies affected the conclusions, and restricted assessment of publication bias in reviews with few included studies. Narrative systematic reviews without quantitative synthesis were generally rated lower than reviews with prespecified meta-analytic methods.

Outcome-level synthesis

Across the included reviews, stem cell–based therapies were associated with possible improvements in neurological impairment, disability, ADL, and motor recovery. However, the magnitude and consistency of the reported effects varied according to cell product, administration route, treatment timing after stroke, study design, and follow-up duration. Detailed review-level estimates are presented in Table 2.

Evidence for infarct volume reduction and biological repair markers was less certain. These outcomes were reported in fewer reviews and were often based on heterogeneous imaging protocols, mixed clinical and preclinical evidence, or limited sample sizes. Therefore, the imaging and biomarker findings were considered exploratory and supportive rather than definitive clinical evidence.

Short-term adverse-event reporting did not show a consistent increase in adverse events or serious adverse events compared with the control groups. However, the available evidence was limited regarding rare or delayed complications, including tumorigenesis, ectopic tissue formation, embolic events, infection, and alloimmunogenicity. Mortality findings were interpreted cautiously because the number of events was small, follow-up durations varied, and several reviews included overlapping primary trials.

Because intervention characteristics varied substantially, the main intervention-level modifiers affecting interpretation are summarized in Table 3. These modifiers included cell source and product type, autologous versus allogeneic origin, dose reporting, passage number and manufacturing characteristics, potency assays, administration route, treatment timing after stroke, follow-up duration, and safety considerations. This summary highlights that MSC-based therapy should not be interpreted as a single uniform intervention.

Table 3.

Intervention-level modifiers affecting interpretation of MSC-based therapy in ischemic stroke.

Modifier	Findings across included reviews	Interpretation for this umbrella review
Cell source/product	Reviews included BM-MSCs, UC-MSCs, AD-MSCs, BM-MNCs, EPCs, PBSCs, MAPCs, neural/progenitor cells, and genetically modified products such as SB623.	Evidence should not be generalized to “MSC therapy” as a single uniform intervention. MSC-focused findings were emphasized, while non-MSC or mixed-cell findings were treated as indirect evidence.
Autologous vs. allogeneic source	Some reviews suggested possible differences between autologous and allogeneic products, but comparative evidence was limited.	No firm conclusion can be made regarding superiority of autologous or allogeneic preparations.
Dose	Dose was variably reported across reviews and was not consistently standardized as cells/kg or total cell number.	Dose–response relationships remain uncertain and require standardized reporting in future trials.
Passage number and manufacturing	Passage number, culture conditions, release criteria, and manufacturing protocols were rarely reported consistently in the included reviews.	Limited manufacturing detail reduces comparability and contributed to downgrading certainty.
Potency assays	Potency measures such as secretome activity, immunomodulatory activity, angiogenic potential, or extracellular vesicle–related assays were not consistently reported.	Lack of standardized potency assays prevents strong conclusions about which MSC product characteristics drive clinical effects.
Route of administration	IV administration was most frequent and appeared more feasible. IA and IC routes may show larger effects in selected contexts but involve greater procedural complexity and risk.	Route-specific interpretation is required; findings should not be pooled conceptually as a single treatment strategy.
Timing after stroke	Reviews included acute, subacute, and chronic phases. Some analyses suggested greater functional gains in subacute or chronic windows.	Timing may modify treatment response, but optimal timing remains unconfirmed.
Follow-up duration	Many studies had short or intermediate follow-up; long-term safety follow-up was limited.	Evidence is insufficient to exclude rare or delayed risks. Long-term surveillance is needed.
Safety considerations	Short-term adverse events were not consistently increased, but rare risks such as tumorigenesis, ectopic tissue formation, alloimmunogenicity, embolic events, and infection were underpowered.	Short-term safety appears acceptable, but long-term safety remains uncertain.

AD-MSC: adipose-derived mesenchymal stem cell; BM-MNC: bone marrow mononuclear cell; BM-MSC: bone marrow–derived mesenchymal stem cell; EPC: endothelial progenitor cell; IA: intra-arterial; IC: intracerebral; IV: intravenous; MAPC: multipotent adult progenitor cell; MSC: mesenchymal stem/stromal cell; PBSC: peripheral blood stem cell; UC-MSC: umbilical cord–derived mesenchymal stem cell; SB623: genetically modified BM-MSC product (SB623 line).

Overlap among included reviews

Across the 26 included reviews, 339 primary-study occurrences were mapped to 111 unique primary studies or trial records. The overall CCA was 8.2%, indicating moderate overlap. Several primary trials or datasets appeared repeatedly across reviews; the detailed normalized overlap mapping is provided in Supplementary Table S1. This overlap was considered in the GRADE certainty assessment and contributed to the decision not to perform a de novo pooled meta-analysis across reviews.

Evidence certainty

The certainty of evidence was assessed at the umbrella-review level using GRADE principles (Table 4). Overall, the certainty of evidence was low to moderate for most neurological and functional efficacy outcomes. Certainty was downgraded because of heterogeneity in cell products, administration routes and treatment timing, small early-phase trials, inconsistency across reviews, imprecision, and overlap among primary studies.

Table 4.

Summary of findings and GRADE certainty for stem cell–based therapies in ischemic stroke.

Outcome	Direction of reported evidence	Certainty	Main reasons for certainty rating	Interpretation
Neurological impairment, NIHSS	Possible improvement in several reviews	Low to moderate	Heterogeneity, small trials, overlap among reviews, variable timing and cell products	Evidence suggests possible neurological benefit, but confidence remains limited.
Disability, mRS	Possible improvement in some reviews	Low to moderate	Inconsistency across reviews, mixed-cell types, variable follow-up, overlap	Findings suggest possible reduction in disability, but estimates are not uniform across reviews.
Activities of daily living, BI/MBI/FIM	Possible improvement	Low to moderate	Heterogeneous scales, route/timing differences, overlapping primary trials	Functional independence may improve, but evidence remains heterogeneous.
Motor recovery, FMA/FMS	Possible improvement	Low	Smaller evidence base, imprecision, inconsistent reporting	Motor recovery findings should be considered preliminary.
Infarct volume/repair markers	Uncertain supportive signal	Very low to low	Limited imaging data, indirectness, mixed clinical/preclinical evidence	Imaging and biomarker findings should be considered exploratory.
Mortality	Possible reduction reported in some pooled analyses, but uncertain	Low	Rare events, variable follow-up, overlap among reviews, early-phase trials	Mortality reduction should be interpreted cautiously and not considered confirmed.
Short-term adverse events	No consistent increase	Moderate	Generally consistent short-term reporting, but limited sample size	Short-term safety reporting is reassuring but not definitive.
Tumorigenesis, ectopic tissue formation, infection, embolic events, alloimmunogenicity	No clear signal reported	Very low to low	Rare events, limited long-term surveillance, underpowered evidence	Absence of reported events should not be interpreted as proof of long-term safety.

NIHSS: National Institutes of Health Stroke Scale; mRS: modified Rankin Scale; BI: Barthel Index; MBI: modified Barthel Index; FIM: functional independence measure; FMA: Fugl-Meyer assessment; FMS: Fugl-Meyer Score; GRADE: Grading of Recommendations Assessment, Development and Evaluation.

The certainty of evidence for motor recovery was judged to be low because of the smaller evidence base and inconsistent reporting. Evidence for infarct volume and repair-related markers was judged to be very low to low because of indirectness, limited imaging data, and mixed clinical and preclinical sources. Evidence for short-term adverse events was judged to be moderate, whereas evidence for rare or delayed safety outcomes, including tumorigenesis, ectopic tissue formation, infection, embolic events, and alloimmunogenicity, was judged to be very low to low. No efficacy outcome or rare long-term safety outcome was judged to have high certainty at the umbrella-review level.

The Summary of Findings table presents the direction of the reported effects, representative estimates extracted from the included reviews, where available, the GRADE certainty ratings, and the main reasons for downgrading. This table should not be interpreted as a de novo pooled analysis. No new pooled effect estimates were calculated in this umbrella review.

Discussion

This umbrella review synthesized 26 systematic reviews and meta-analyses evaluating stem cell–based therapies for ischemic stroke, with an emphasis on MSC-based interventions. The evidence suggests possible improvements in selected neurological and functional outcomes; however, the strength of inference remains limited by review overlap, heterogeneous interventions, small early-phase trials, variable follow-up durations, and limited long-term safety surveillance.^{6–8,10,17,22,23,26,28,30,35}

The most consistent clinical signals involved neurological impairment, disability, ADL, and motor recovery. However, these signals should not be interpreted as definitive evidence of efficacy. The included reviews differed in outcome measures, follow-up durations, trial designs, cell products, and analytical methods. In addition, many reviews included the same primary studies, meaning that concordant findings across reviews may partly reflect repeated use of the same evidence base rather than independent confirmation.

The route and timing of administration may modify treatment effects, but the current evidence remains hypothesis-generating. Intravenous administration was the most frequently studied route and may be more feasible for large-scale clinical testing. Intra-arterial and intracerebral routes may show larger point estimates in selected analyses, but they involve greater procedural complexity and potential risk. Similarly, some subgroup analyses suggested possible differences among acute, subacute, and chronic treatment windows, but few trials were designed to compare treatment windows directly. Therefore, no administration route or treatment window should be considered definitively superior based on the current evidence.^{8,10,17,18,26,28}

Cell-product heterogeneity is another important limitation. Reviews included bone marrow mononuclear cells (BM-MSCs), umbilical cord mesenchymal stem cells (UC-MSCs), adipose-derived mesenchymal stem cells (AD-MSCs), bone marrow mesenchymal stem cells (BM-MNCs), endothelial progenitor cells (EPCs), peripheral blood stem cells (PBSCs), multipotent adult progenitor cells (MAPCs), neural/progenitor cell products, and genetically modified cell products. Dose, passage number, culture conditions, release criteria, and potency assays were not consistently reported. These differences limit the ability to identify dose–response relationships or determine which biological characteristics are responsible for the observed clinical effects. Accordingly, MSC-based therapy should not be considered a single, uniform intervention, and future interpretation should remain product-specific.

Short-term safety reporting was generally reassuring, with no consistent increase in adverse events or serious adverse events. However, the current evidence is insufficient to exclude rare or delayed risks, such as tumorigenesis, ectopic tissue formation, alloimmunogenicity, infection, embolic events, or procedure-related complications. Mortality reduction was reported in some pooled analyses, but this finding remains uncertain because the number of events was small, follow-up durations varied, and overlap among the primary studies was present. Therefore, mortality findings should be interpreted as an uncertain signal requiring confirmation rather than as evidence of a survival benefit.^6,14,28,30

The overlap assessment was central to the interpretation of the evidence. Across the 26 reviews, 339 primary-study occurrences were mapped to 111 unique primary studies or trial records, with an overall CCA of 8.2%, indicating moderate overlap. This overlap supports a conservative interpretation of repeated findings and justifies the decision not to conduct a de novo pooled meta-analysis across reviews, which could have double counted participants and overestimated precision. ³⁶

Overall, the current evidence supports continued clinical investigation rather than routine clinical implementation. Future trials should use standardized cell-product characterization, harmonized outcome measures, prespecified comparisons of administration route and treatment timing, and long-term safety monitoring. Until adequately powered multicenter trials confirm efficacy and long-term safety, MSC-based therapy for ischemic stroke should remain an investigational approach.

Strengths

This umbrella review followed a prospectively registered protocol, adhered to the PRISMA 2020 guidelines and JBI recommendations, and applied dual independent screening, data extraction, and appraisal. We used AMSTAR 2 to assess methodological quality and GRADE principles to summarize the certainty of evidence.^11–13 The review included 26 systematic reviews and meta-analyses published between 2011 and 2025 and remained focused on evidence from human studies of ischemic stroke. Another strength is that the findings were organized according to clinically relevant outcome domains and interpreted with attention to cell product, administration route, treatment timing, safety, and overlap among primary studies. ³⁶

Limitations

This umbrella review has several limitations. First, substantial heterogeneity was present across the underlying primary trials and reviews, including differences in cell source, donor type, manufacturing protocols, dose, passage number, potency assessment, administration route, timing after stroke, rehabilitation background, and outcome assessment windows. Second, overlap among the included reviews may have inflated the apparent consistency of the evidence because several reviews included many of the same primary trials. This overlap was considered in the certainty assessment and was one reason that no de novo pooled meta-analysis was performed. ³⁶ Third, many primary studies were small, early-phase trials or were at risk of bias, limiting confidence in the efficacy estimates. Fourth, long-term safety evidence was limited; therefore, rare or delayed risks, such as tumorigenesis, ectopic tissue formation, alloimmunogenicity, infection, and embolic events, cannot be excluded. Fifth, the restriction to English-language full-text reviews may have introduced language bias and excluded relevant non-English evidence. Sixth, several included reviews combined MSCs with other stem cell or progenitor cell products; although the findings were interpreted according to intervention category where possible, residual indirectness remains. Finally, data on real-world comparative effectiveness, cost-effectiveness, scalability, manufacturing feasibility, and health system implementation remain limited.

Future directions

Future research on MSC-based therapy for ischemic stroke should prioritize standardization, adequate statistical power, and long-term safety surveillance.

First, future trials should use clearly defined CMC criteria, including cell identity, donor source, viability, dose, passage number, culture conditions, release criteria, and potency assays. Potency assessment should include relevant biological functions, such as immunomodulatory activity, angiogenic potential, neurotrophic signaling, and secretome or extracellular vesicle activity. Because extracellular vesicle- and secretome-based measures can be affected by storage temperature, freezing protocols, freeze-thaw cycles, stabilizers, transport conditions, and storage materials, release and storage standards should be incorporated into future protocols. ³⁷

Second, adequately powered multicenter RCT are needed to compare leading cell products, administration routes, and treatment timing strategies. Intravenous delivery may be the most feasible route for large-scale testing, whereas intra-arterial and intracerebral delivery should be evaluated only in settings where procedural risk and patient selection can be carefully controlled. Acute, subacute, and chronic treatment windows should be analyzed separately or compared using prespecified subgroup designs.^{8,17,18,26,28}

Third, future trials should adopt harmonized core outcome sets with standardized assessment windows. Primary outcomes should include validated neurological and functional measures, such as the NIHSS, mRS, BI/modified Barthel Index (MBI), FMA, and FIM. Secondary outcomes may include quality of life, patient-reported outcomes, mortality, imaging markers, and biomarkers related to angiogenesis, neurotrophic signaling, and inflammation.

Fourth, long-term safety surveillance is essential. Future studies should include active follow-up for at least 24–60 months, where feasible, with systematic monitoring for tumorigenesis, ectopic tissue formation, alloimmunogenicity, infection, embolic events, vascular complications, and route-related adverse events. These outcomes should be reported even when no events occur.

Finally, future evidence syntheses should account for overlap among primary studies. Individual patient data meta-analysis, prospective meta-analysis, and updated network meta-analysis using uniform effect measures may help identify treatment-effect modifiers and reduce the risk of duplicated evidence. At present, routine clinical implementation should await adequately powered multicenter trials with harmonized outcome assessment and long-term safety monitoring.

Conclusions

The evidence suggests that stem cell–based therapies, particularly MSC-based approaches, may improve selected neurological and functional outcomes after ischemic stroke. However, confidence in their efficacy remains limited by overlapping reviews, heterogeneous cell products, variable administration routes and treatment timing, small early-phase trials, and limited long-term safety follow-up. Short-term safety reporting is generally reassuring, but the available evidence remains insufficient to exclude rare or delayed complications. Overall, the current evidence supports further standardized, adequately powered, multicenter trials with harmonized outcome measures and long-term safety surveillance rather than routine clinical implementation at this stage.