The present systematic review was prospectively registered in PROSPERO (CRD420251236846) under the title ‘Mesenchymal Stromal Cells and MSC-Based Cell-Free as Acute Traumatic Brain Injury Therapy: Systematic Review and Meta-Analysis of preclinical studies’. The protocol is publicly available at https://www.crd.york.ac.uk/PROSPERO/view/CRD420251236846. All methods were pre-defined in the registered protocol to minimize reporting bias and ensure methodological transparency. Any deviations from the protocol will be clearly documented in the final publication.

A systematic literature search was conducted on November 21, 2024 across four major electronic databases: PubMed, MEDLINE, Scopus and Cochrane Library. The search combined controlled vocabulary (MeSH) and free-text keywords with Boolean operators. The search utilized a combination of MeSH terms and free-text keywords as demonstrated in Table I.

To ensure comprehensive coverage, reference lists of retrieved articles were manually screened through backward and forward citation searching for additional relevant studies. To capture contemporary evidence, the search was restricted to studies published over the past decade (January 1, 2015 to the search date) with no language restrictions. The review was written and published in English.

Studies were included if they met all of the following: Investigated MSCs and/or MSC-derived exosomes (MSC-derived cell-free products, conditioned medium) administered ≤7 days post-injury (defined as the acute phase, aligned with the secondary injury cascade in TBI); pre-clinical in vivo animal models of TBI (any severity or model); and reported quantitative data for at least one pre-defined efficacy outcome (Fig. 1). Comparators comprised vehicle, sham, saline, or standard care. Both randomized and non-randomized study designs were included. The timing of administration was extracted from the Methods section of each study and independently verified by two reviewers (DRA and AP) to ensure compliance with the ≤7-day criterion. Articles not meeting these criteria were excluded (non-traumatic brain injury models, in vitro-only studies and reviews).

The initial search yielded a broad range of articles, which were filtered based on publication year, relevancy and study type. Of note, two reviewers (C and TS) independently screened the titles and abstracts of all records identified through the systematic search and manual reference checking. Full-text reports of any record considered potentially relevant by at least one reviewer (C) were retrieved and independently evaluated for inclusion by the same two reviewers (DRA and AP). Any discrepancies between reviewers (DRA and AP) during title/abstract screening or full-text assessment were first resolved by discussion. Persistent disagreements were adjudicated by a third reviewer (C). Exclusions at the full-text stage were categorized and recorded. A PRISMA flow diagram was created to illustrate the study selection process (Fig. 1).

Selected articles with key data were systematically extracted, including the title and authors of the articles, year of publication, study objectives and methodology, sample size and study population (animal species demographics), intervention details (MSC and/or exosome source, route of administration), comparators, and key findings related to efficacy and safety outcomes. Primary outcomes were the following: i) Neurological function: Assessed by the modified neurological severity score (mNSS; score 0-18, lower scores indicate better function); ii) cognitive performance: Evaluated using the Morris water maze (MWM) test (typically reported as mean escape latency in seconds or percentage time spent in the target quadrant); iii) histopathological outcome: Lesion volume (mm³), hematoxylin and eosin staining or magnetic resonance imaging (MRI).

Data were extracted independently by two reviewers (C and TS) using a standardized Excel form. Disagreements were resolved by discussion or a third reviewer (C). Extracted items included study details, intervention characteristics, sample sizes and results. Continuous outcomes data were used directly or converted as the mean and SD. Graphs were digitized using WebPlotDigitizer (v4.5) when needed and sensitivity analyses tested earliest or all time points. Data synthesis was conducted using RevMan Web, aggregating quantitative data for the meta-analysis where applicable, using random-effects models. Heterogeneity was assessed with I². Authors were contacted to provide any required data not available in published reports.

Extracted data were analyzed to identify recurring themes, magnitude of treatment effects, optimal therapeutic parameters and research gaps. Similarities and differences across pre-clinical and clinical studies were evaluated to assess the translational potential of MSCs and/or MSC-derived exosomes administered within 7 days following TBI. Meta-analytic techniques included the calculation of pooled effect sizes with 95% confidence intervals (CIs) for continuous outcomes (standardized mean difference) and dichotomous outcomes (risk ratio), with heterogeneity assessed using the I² statistic. Subgroup analyses based on study type, intervention (MSC vs. exosome), timing, dose, route and TBI severity, along with funnel plot assessments for publication bias, were conducted to strengthen the validity of conclusions. The certainty of findings was assessed using the Cochrane RoB Tools 2.0 checklist to assess the risk of bias within the studies included in the present systematic review. The assessment was integrated into the systematic review and meta-analysis process, which adheres to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) protocols and The Cochrane Handbook. This approach ensures a rigorous evaluation of the evidence, aligning with the standards of high-quality systematic reviews that employ established criteria to determine confidence in the body of evidence for key outcomes.

Risk of bias was assessed independently by two reviewers (C and TS) using SYRCLE's tool for animal studies. Domains were rated low, high, or unclear risk. Disagreements were resolved by consensus or a third reviewer (C). Certainty for primary outcomes (mNSS, MWM, lesion volume) was evaluated using an adapted GRADE approach for pre-clinical studies and the results are shown in Table SI. Beginning from ‘high’, downgrades were applied for risk of bias (SYRCLE), inconsistency (high I²), indirectness, imprecision and publication bias (funnel plots/Egger's test).

The key characteristics of the included interventions are summarized in Fig. 2. Among the 46 pre-clinical studies, a greater proportion of experiments used MSC-based cell therapy than MSC-derived cell-free products (59 vs. 41%; Fig. 2A). As regards the delivery route, almost half of the interventions were administered IV (48%), followed by ICV injection in 40% of the studies (Fig. 2B). Only a small minority used alternative routes, such as intranasal delivery (6%) or other less frequently applied approaches (intra-arterial, intraperitoneal and intra-retrobulbar; ≤2% each), indicating that the current pre-clinical evidence is dominated by systemic IV and ICV administration strategies.

The SYRCLE risk-of-bias assessment across all 46 included pre-clinical studies is summarized in Fig. 3. Overall, the majority of studies had at least one domain with an unclear or high risk of bias. In total, 43% of the assessments were rated as low risk, 53.5% as unclear risk and 3.5% as high risk. Domains related to random sequence generation, allocation concealment, random housing and blinding of caregivers or investigators were predominantly judged as unclear risk due to insufficient reporting. By contrast, the domains of incomplete outcome data, selective reporting and other sources of bias were mostly assessed as low risk.

The visual inspection of funnel plots for the three main outcomes is illustrated in Fig. 4. For mNSS (Fig. 4A), the funnel plot appeared mildly asymmetric, with a relative paucity of small, imprecise comparisons reporting null or detrimental effects and a cluster of small studies favoring MSC-based therapies, suggesting the presence of small-study effects and possible publication bias. A similar pattern of moderate asymmetry was observed for lesion volume (Fig. 4C), where several imprecise studies reported very large reductions in lesion size without a corresponding number of small studies showing neutral effects. By contrast, the funnel plot for MWM (Fig. 4B) was more symmetric around the pooled effect line, providing no strong visual evidence of publication bias. Nevertheless, given the high between-study heterogeneity and the experimental nature of the included studies, all pooled estimates should be interpreted with caution, particularly for mNSS and lesion volume.

The meta-analysis was then stratified by outcome domain. For global neurological function, 92 comparisons reporting mNSS were pooled. Using a random-effects model with the Hartung-Knapp adjustment, MSC-based and MSC-derived cell-free therapies significantly reduced mNSS scores compared with the control (95% CI; P<0.0001; I²=100%), indicating improved neurological recovery. The corresponding pooled and stratified effects are presented in Fig. 5.

For cognitive performance, 51 comparisons reporting MWM outcomes were combined. The pooled analysis demonstrated substantially lower (improved) MWM scores in the treatment groups than in the controls [overall mean difference (MD), -16.72; 95% CI, -22.87 to -10.58; P<0.00001; I²=100%], these findings are presented in Fig. 6. For structural brain damage, 50 comparisons reporting lesion volume were analyzed and a robust reduction in lesion volume was observed in the treated animals compared with the controls (overall MD, -0.15; 95% CI, -0.17 to -0.14; P<0.00001; I²=98%) (Fig. 7). Overall, across neurological deficit (mNSS), cognition (MWM) and lesion volume, stratified analyses consistently favored MSC-based or MSC-derived cell-free therapies in pre-clinical TBI models, albeit with considerable between-study heterogeneity.

Subgroup analyses were then performed according to product type (MSC vs. MSC-derived cell-free products) and route of administration (ICV, IV or other routes). For mNSS, subgroup analyses suggested that the magnitude of benefit varied across intervention types (test for subgroup differences: P=0.02; I²=60%). As shown in Fig. 5, larger pooled improvements for MSC-derived cell-free products were observed in mNSS (MD, ~-4.7) compared with MSC-based cellular therapies, while the ICV delivery of cell-free products yielded very imprecise estimates with wide confidence intervals due to the small number of experiments. Despite this variability, all subgroups favored treatment over the control.

For MWM, consistent cognitive benefits were observed across subgroups as shown in Fig. 6. Both MSC-based and cell-free interventions, delivered either ICV or IV, improved MWM performance to a similar extent, and the test for subgroup differences was not significant (P=0.78; I²=0%). This finding suggests that, for cognitive recovery, the beneficial effect is relatively robust to the choice of MSC product type and delivery route.

For lesion volume, subgroup analyses revealed clearer differences between modalities (test for subgroup differences: P<0.00001; I²=95.7%), As shown in Fig. 7, MSC-derived cell-free therapies produced the largest reductions in lesion volume, particularly when delivered intravenously (pooled MD, -0.21 to -0.24), whereas MSC-based cellular therapies exhibited, minimal, but still significant effects (pooled MD, -0.09). Nevertheless, each subgroup exhibited a shift in favor of treatment compared with the control, indicating that both MSC and MSC-derived cell-free approaches confer structural neuroprotection, with cell-free products, particularly via IV administration, tending to provide the greatest lesion-reducing effect.

Taken together, these subgroup findings support a consistent neuroprotective signal across different MSC-based strategies, while also suggesting that MSC-derived cell-free products and IV delivery may be particularly promising for optimizing functional and structural outcomes after experimental TBI.

All primary outcomes exhibited high heterogeneity (I²=98-100% for mNSS, lesion volume and MWM), typical in pre-clinical TBI meta-analyses due to variations in animal species, TBI models, injury severity, MSC/exosome sources, isolation methods, doses, administration routes (IV vs. ICV) and follow-up durations. Sensitivity analyses (leave-one-out analysis) confirmed that pooled estimates remained directionally consistent and significant in favor of treatment, with no single outlier dominating the results (Figs. 5, 6 and 7). Pre-specified subgroup analyses by product type (MSC vs. cell-free) and route (IV vs. ICV vs. other) partially accounted for heterogeneity; notably, IV cell-free interventions showed larger effects on lesion-reducing effect (subgroup difference P<0.05 for lesion volume). Residual heterogeneity remained high within subgroups, likely due to unmeasured factors such as exosome characterization and preconditioning.

Meta-regression was not conducted owing to insufficient studies per covariate and a very high baseline heterogeneity. The most robust conclusion is the consistent directional benefit across diverse pre-clinical settings: MSC-based therapies and, particularly IV. MSC-derived (exosomes) demonstrate clear neuroprotective signals in neurological function, cognition, and lesion volume. These results underscore paracrine mechanisms and emphasize the urgent need for standardized protocols in future pre-clinical studies to facilitate clinical translation.

In the present meta-analysis, 403 animals contributed mNSS data (194 experimental and 209 controls). Across all product types and routes, MSC-based interventions significantly improved global neurological function following experimental TBI, as shown by a lower mNSS in treated animals (P<0.0001). This indicates a consistent neuroprotective effect on composite motor, sensory, reflex and balance deficits, and aligns with pre-clinical studies in which systemic MSC administration upregulated neurotrophic factors [brain-derived neurotrophic factor (BDNF), vascular endothelial growth factor (VEGF) and nerve growth factor (NGF)] in the peri-lesional cortex, reduced apoptosis, and improved behavioral recovery (12).

The overall neuroprotective effect is largely attributable to paracrine mechanisms, whereby MSCs secrete anti-inflammatory cytokines, neurotrophic factors and extracellular vesicles that modulate the post-injury microenvironment without requiring long-term cellular engraftment. MSC-derived exosomes, as key paracrine mediators, carry bioactive cargo including proteins (BDNF and VEGF for neuroprotection and angiogenesis) and miRNAs (miR-133b for axonal growth and synaptic plasticity, miR-216a-5p for inhibiting neuroinflammation via BDNF pathways, miR-124-3p for suppressing Rela/Apolipoprotein E to mitigate neurodegeneration). Recent studies (2024-2025) confirm that these cargoes enable exosomes to cross the blood-brain barrier efficiently, deliver targeted payloads to neurons and glia, and regulate transcriptional/translational processes for repair (67,68).

Exosomes shift microglial polarization from pro-inflammatory M1 (iNOS, TNF-α, IL-1β and IL-6) to reparative M2 phenotype (Arg-1, CD206 and IL-10), downregulating NF-κB/MAPK pathways and reducing oxidative stress/apoptosis (68,69). The study by Liu et al (67) highlighted exosomal miR-133b and miR-22 upregulation under hypoxic conditions to facilitate nerve repair, while Xiong et al (68) emphasized neurorestoration via anti-apoptotic and anti-inflammatory cargo. These mechanisms explain greater mNSS improvements with cell-free products, particularly IV exosomes, as they provide concentrated, standardized delivery without cell survival issues (8,10-12). Beyond immunomodulation, both MSCs and MSC-derived exosomes attenuate neuronal apoptosis, promote angiogenesis, support neurogenesis and enhance synaptic plasticity, all of which contribute to the restoration of cortical and subcortical networks that mediate motor and sensory functions (11,12). These effects align with the longstanding view that MSCs exert their therapeutic effects predominantly through paracrine signaling rather than direct cell replacement, a concept supported by accumulating pre-clinical and early clinical evidence in TBI (13,70).

However, high heterogeneity (I²=100%) substantially limits the precise interpretation of pooled estimates. Sources include variability in TBI models controlled cortical impact (CCI) and fluid percussion injury (FPI), injury severity, MSC sources (bone marrow vs. umbilical cord), exosome isolation/characterization methods, doses, timing (≤7 days), routes (IV vs. ICV) and follow-up durations. This heterogeneity reflects real-world translational challenges, but also underscores the robustness of directional benefits across diverse settings (70). Future studies are required to standardize protocols Minimum Information for Studies of Extracellular Vesicles (MISEV) guidelines for exosome characterization, consistent dosing in particle number or protein content) to reduce variability and facilitate meta-regression or clinical trial design. Sensitivity analyses confirmed directional consistency, but residual heterogeneity within subgroups suggests unmeasured factors (preconditioning, miRNA cargo profiling) as key contributors.

Although almost all point estimates favored treatment over control, statistically significant gains were driven mainly by MSC-derived cell-free preparations, particularly when delivered intravenously, whereas MSC subgroups exhibited similar directions, but wider confidence intervals. This pattern suggests that IV cell-free products, particularly exosomes, may provide larger and more reliable benefits, likely due to better bioavailability and targeted immunomodulation.

For cognitive outcomes, 408 animals with MWM data were included. Cognitive recovery is a central target for TBI therapies, and tje pooled MWM analysis revealed that MSC-based interventions significantly improved spatial learning and memory compared with the controls, despite marked between-study heterogeneity (P<0.00001; I²=100%). This indicates that, at a global level, MSC-centered strategies do not only reduce gross neurological deficits, but also confer measurable benefits on higher-order functions that rely on hippocampal integrity and neuroplasticity.

Both MSCs and MSC-derived cell-free products act on multiple levels of hippocampal circuitry. Pre-clinical TBI research has demonstrated that intravenous MSC-derived exosomes enhance spatial learning by increasing dentate gyrus neurogenesis (BrdU⁺/DCX⁺ and BrdU⁺/NeuN⁺ cells), promoting synaptic plasticity (upregulation of GAP-43, synaptophysin and PSD-95), and reducing neuronal loss in CA1-CA3, changes that parallel improved MWM performance (50). BDNF-induced MSC exosomes further augment these effects by delivering miR-216a-5p and other neurotrophic cargos that support neuronal survival and dendritic complexity, thereby accelerating recovery of spatial learning following TBI (13). In parallel, hypoxia-preconditioned MSCs promote oligodendrogenesis and remyelination, restore white-matter integrity, and activate mTOR/HIF-1α/VEGF signaling, which together improve network connectivity and result in better cognitive outcomes in MWM testing (42).

Cell-free preparations appear particularly effective as they concentrate the paracrine signals that drive hippocampal plasticity. MSC-derived exosomes modulate microglial phenotypes in the hippocampus, shifting from pro-inflammatory M1 to anti-inflammatory M2 via miR-181b/IL-10-STAT3 signaling, leading to reduced IL-1β, IL-6 and TNF-α, less synaptic pruning, and improved MWM performance (13,52,54). Other exosomal cargos, such as the long non-coding RNA MALAT1 have been shown to activate pro-regenerative and synaptogenic pathways, and to attenuate chronic neuroinflammation when delivered intranasally following TBI, again translating into improved motor and cognitive scores (62). Together with broader evidence that MSC-based therapies enhance neurogenesis, angiogenesis and network-level plasticity across neurological models, these data provide a biological rationale for the finding of the present study that the majority of MSC-based and MSC-derived cell-free interventions improve spatial learning and memory, with the most robust and consistent cognitive gains arising from standardized exosome-based strategies (71).

Subgroup analyses revealed that almost all point estimates favored treatment, indicating a broadly consistent trend toward better MWM performance with both MSC and MSC-derived cell-free products. However, statistically significant improvements were largely driven by the cell-free subgroup (P<0.00001), whereas MSC subgroups (overall MSC, MSC-ICV and MSC-IV) exhibited effects in the same direction but failed to reach conventional significance (all P>0.05), reflecting wide confidence intervals and substantial heterogeneity. The test for subgroup differences was not significant (P=0.78, I²=0%), meaning that confidence intervals overlapped and formal interaction testing does not prove clear superiority of one modality over another. Practically, these findings suggest that the majority of MSC-based and cell-free strategies tend to improve cognitive performance, with the strongest and most statistically robust signal arising from cell-free interventions. This aligns with the concept that post-TBI cognitive recovery depends heavily on synaptic remodeling, neurogenesis and network-level plasticity driven predominantly by the paracrine cargo of MSC-derived secretome and extracellular vesicles, while the very high overall heterogeneity and variable methodological quality emphasize the need for more standardized MWM protocols, predefined treatment timing, and rigorous blinding in future pre-clinical studies.

For structural outcomes, 403 animals contributed lesion volume data, with 194 animals in the experimental groups and 209 in the control groups. Lesion volume provides a structural correlate of tissue preservation and is closely linked to long-term functional outcome following TBI. The meta-analysis demonstrated a very robust and highly significant reduction in lesion volume in animals treated with MSC-based interventions compared with controls (P<0.00001, I²=98%). This magnitude of effect indicates that MSC-centered strategies consistently limit the extent of brain tissue loss across diverse experimental models, injury severities, and treatment protocols.

At the subgroup level, a markedly coherent pattern was observed: All subgroups (lesion volume, MSC, MSC-ICV, MSC-IV, cell-free and cell-free IV) exhibited highly significant reductions in lesion size (P<0.001). Thus, in contrast to the more variable statistical significance observed in mNSS and MWM, the structural endpoint of lesion volume exhibited a uniformly strong treatment signal across both cell-based and cell-free approaches. Nevertheless, the test for subgroup differences was highly significant, with very high heterogeneity between subgroups (P<0.00001, I²=95.7%), indicating that the magnitude of neuroprotection is not identical for all modalities.

This structural protection is consistent with pre-clinical evidence that MSCs and their extracellular vesicles primarily act by limiting secondary injury cascades in the peri-lesional ‘penumbra’ rather than reversing the primary mechanical damage (72). MSC-derived exosomes rapidly reach the injured cortex and hippocampus, where they modulate microglia/macrophage polarization from a pro-inflammatory M1 phenotype to a reparative M2 state, downregulating iNOS and pro-inflammatory cytokines, while upregulating Arg-1 and CD206(12). By dampening neuroinflammation and oxidative stress, these vesicles reduce apoptotic cell death in neurons and oligodendrocytes and thereby preserve viable tissue, which translates into smaller cavitary lesions on histology and MRI (13,68).

In parallel, MSCs and MSC-derived extracellular vesicles enhance neurovascular remodeling and white-matter integrity, two additional processes that constrain lesion expansion. Hypoxia-preconditioned bone marrow-derived MSCs promote remyelination and reduce white-matter injury via mTOR/HIF-1α-dependent pathways, leading to reduced overall lesion volume in mice with TBI (42). MSC/EV therapies also stimulate angiogenesis and restore blood-brain barrier integrity, which decreases edema and secondary ischemic damage in the peri-contusional zone (73,74). Complementary data from broader extracellular vesicle research indicate that stem cell-derived exosomes can lower cortical and hippocampal water content and lesion volumes, while protecting against mitochondrial oxidative stress and apoptosis (75).

The findings of the present study are summarized by the illustration in Fig. 8 demonstrating different administration routes of MSCs and exosomes in rodent TBI models (IN, ICV and IV) and their effects on functional, cognitive and structural outcomes. Across all three panels, the bars indicate that MSC-derived cell-free products, particularly when administered IV, tend to produce the greatest improvement compared with whole-cell MSC therapy.

Although the present meta-analysis revealed consistent neuroprotective effects of MSC-based therapies and MSC-derived exosomes in acute pre-clinical TBI models, translation to clinical practice remains challenging, with no registered trials specifically for acute TBI using MSC-derived exosomes as of 2025. Key barriers include the lack of standardization in exosome production, characterization and cargo consistency (MISEV guidelines), undefined optimal dosing, therapeutic window (≤7 days in rodents vs. variable human timing), administration route (IV most promising), and long-term safety concerns (low immunogenicity). To advance translation, future efforts should prioritize large-animal models for better bridging to humans, standardized protocols, multicenter pre-clinical studies with long-term follow-up, and early-phase human trials focused on safety, biodistribution, and biomarkers (73).

In line with previous pre-clinical TBI meta-analyses, high statistical heterogeneity limits precise interpretation of pooled effect sizes, arising from variations in animal species, TBI models, injury severity, MSC/exosome sources, doses, routes, timing and follow-up durations; residual heterogeneity suggests unmeasured factors; thus, estimates should be viewed as directional summaries rather than protocol-specific predictions (73). The risk of bias (SYRCLE tool) was often unclear or high (43% low risk), particularly in randomization, blinding and allocation concealment, potentially inflating positive effects as indicated by funnel plot asymmetry. Translational limitations include fundamental differences between rodent models and human TBI (smaller brain size, distinct neurovascular/immune responses, absence of comorbidities such as age or polytrauma, and species-specific biodistribution/blood-brain barrier penetration), complicating direct extrapolation. Despite these constraints, the consistent directional benefit across settings supports the paracrine neuroprotective potential of MSC and exosome therapies, warranting standardized protocols for improved reproducibility and clinical relevance.

In conclusion, the present systematic review and meta-analysis demonstrates that MSC-centered interventions provide consistent neuroprotective effects in pre-clinical models of traumatic brain injury. Across the included experiments, treatment groups exhibited lower neurological deficit scores, improved performance on cognitive testing, and smaller lesion volumes than controls, indicating improvement at functional, cognitive and structural levels. Stratified and subgroup analyses further suggested that MSC-derived cell-free products, particularly when administered intravenously, tended to yield larger and more precisely estimated benefits than whole-cell MSC preparations, particularly for cognitive recovery and lesion volume reduction. Taken together, these findings support a predominant role of paracrine mechanisms in MSC-mediated neuroprotection and identify intravenously delivered cell-free MSC therapies as the most promising MSC-based strategy in current pre-clinical TBI research.