Extracellular Vesicles Are Associated With Biological, Mechanical, and Structural Tendon Healing Metrics in Preclinical Models: A Systematic Review and Meta‐Analysis

ABSTRACT

Tendinopathies comprise half of musculoskeletal injuries, yet medical treatments are lacking. Orthobiologics are novel therapeutics developed to address this gap, with extracellular vesicles (EVs) being a promising prospect. However, we lack a complete understanding of how EVs affect regeneration, both endogenously and as a tool for augmenting tendon healing. Therefore, the purpose of this review was to compile the literature focused on understanding the impact of EV treatment on tendon healing using animal models. We performed a literature search in accordance with PRISMA reporting guidelines. Title, abstract, and full‐text screening were performed by two authors. We assessed transparency and risk‐of‐bias using the ARRIVE, MISEV, and SYRCLE Guidelines, and meta‐analyses were performed on overlapping outcome variables. After screening, 46 articles were included. Most studies utilized young male Achilles tendons, with a diversity of EV cell sources, injury types, dosage schemes, and biomaterials. Several ARRIVE (e.g., blinding) and MISEV (e.g., seeding density) recommendations remain poorly reported in this literature, and the risk of bias per SYRCLE guidelines is relatively high. Despite high heterogeneity, EV treatment was associated with improved histological tendon healing scores and increased load to failure, tendon stiffness, and Young's modulus. Further meta‐analyses revealed EV treatment was associated with increased type I collagen protein and type III collagen transcript expression, and decreased protein expression of IL‐6 and markers of senescence. We recommend that future studies include different ages, sexes, and more systematically protocolized EV delivery scheme.

Statement of Clinical Significance: This systematic review identifies gaps in the tendon and EV literature while also unveiling unexplored mechanisms by which EVs may promote tendon healing. This improved understanding and call for rigor in studies enhances the translational potential of EVs as a future treatment for tendon injuries.

1. Introduction

Approximately 50% of musculoskeletal injuries involve tendons [1, 2]. Specifically, the incidence of acute tendon ruptures in the United States is estimated to be 3.80 per 100,000 person‐years [3], and 22% of adults over the age of 60 are estimated to have chronic tendon tears [4, 5]. Currently, activity modification, physical therapy, and anti‐inflammatory medications are the mainstays of non‐surgical interventions [6]. However, 10%–45% of tendinopathies fail conservative treatment [6]. Of these patients, 29%–42% ultimately have surgical intervention [7], with re‐tear rates up to 94% [8, 9]. For patients who opt for non‐surgical management, 35%–50% may continue to have chronic symptoms [10, 11, 12]. Thus, there is a significant need for improved medical treatment of tendon pathology.

Over the last two decades, clinicians and scientists have explored novel orthobiologic interventions for treating tendinopathies. Currently, platelet rich plasma (PRP) [13], bone marrow aspirate concentrate (BMAC), micro‐fragmented adipose tissue (M‐FAT) [14], and collagen patch augmentation are increasingly utilized in the clinical settings [15]. However, the efficacy of such treatments has been challenging to determine as we lack an understanding of how these orthobiologics contribute to tendon healing, if at all. These shortcomings are due, at least in part, to research studies being consistently plagued by heterogenous design, inconsistent outcomes, and incomplete reporting [13, 15].

Extracellular vesicles (EVs) have the potential to both close the gap in our mechanistic understanding of native tendon regeneration and provide an orthobiologic tool to enhance healing. EVs are nanometer to micrometer sized membranous structures containing nucleic acids, proteins, and lipids [16]. They are secreted by nearly all cells and tissues and have increasingly been shown to play a role in healing through recruiting inflammatory cells to the injury site, modulating resident cell phenotype, increasing proliferation, and promoting tissue repair [17]. As an acellular biologic, EVs have low immunogenicity, may have better stability in long term storage, and can be engineered to stimulate cell specific reactions [18, 19], suggesting they may enable more individualized, precision medicine therapeutic strategies in comparison to current cell‐based approaches [20].

However, similar to other orthobiologics, high variability in experimental design among pre‐clinical studies remains a consistent challenge. Donor source, isolation and characterization methods, as well as dosage and frequency of EV administration differ widely, in addition to the tendon and injury model utilized. In the last 5 years, several reviews have evaluated how EVs modulate tendon‐bone incorporation [21], Achilles tendon healing [22], and the tendon immune environment [23]. Systematic reviews of the pre‐clinical literature focused on understanding how EVs impact tendon healing are limited [21, 22, 24], none have utilized meta‐analyses to identify possible mechanisms, and assessment of transparency in reporting have been limited. Therefore, our goal was to systematically review the published literature, quantify transparency in reporting and risk‐of‐bias among said literature using published guidelines, and meta‐analyze how EV administration impacts tendon healing following injury using in vivo animal models. In line with the 2023 Minimal Information for Studies of Extracellular Vesicles (MISEV) reporting guidelines, we define EVs as particles naturally released from cells, delimited by a lipid bilayer, and unable to replicate independently [25]. This generic, umbrella term includes exosomes, small EVs (< 200 nm), medium/large EVs (> 200 nm), and apoptotic vesicles.

2. Methods

2.1. Systematic Review

We performed a systematic review of the literature in accordance with the Preferred Reporting Items for Systematic Reviews and Meta‐Analyses (PRISMA) reporting guidelines (Appendix S1) [26], Meta‐Analysis of Observational Studies in Epidemiology (MOOSE) checklist (Appendix S2) [27], Cochrane Handbook for Systematic Reviews of Interventions [28], and the practical guide for meta‐analysis in animal studies [29]. Although we did not register this systematic review on PROSPERO a priori [30], we followed a pre‐specified protocol for inclusion and exclusion of articles, as well as for our analyses. We identified eligible manuscripts using the following PECO [31]: Population–Animals (no species, age, or sex limitation), Exposure—EV treatment following an injury (no limitation on type of native EV, where the EVs came from, or injury type), Comparison—no EV treatment following an injury (age/sex/injury matched control), and Outcome–metrics of tendon health and healing. Specific search terms are listed in Table S1. Studies were excluded if they were written in a non‐English language, not peer reviewed, not original research articles, did not include age/sex matched controls, there was no tendon injury performed in either group, or the “injury” was an infection. There was no exclusion based on the year of publication. Studies that exclusively included engineered EVs (e.g., the experimental group was engineered EVs and the control group were native EVs) were excluded. If a study contained a native EV, an engineered EV group, and a control, only the native EV group was included as an experimental group. We elected to exclude engineered EVs from this analysis to ensure comparability of intrinsic EV bioactivity across studies.

On April 13, 2026, we performed a literature search on PubMed, Physiotherapy Evidence Database (PEDro), EBSCO, Web of Science, Scopus, and Cochrane Central Register of Controlled Trials in attempt to exhaustively collect all preclinical articles on this topic. Two authors independently screened the titles, abstracts, and full texts of studies identified via our search based on the aforementioned PECO and Cochrane Handbook recommendations [28]. Reasons for exclusion were documented at the full text level (Table S2). Following full text screening, we performed a citation search via PubMed to identify additional publications for screening. Screening was completed in pre‐designed spreadsheets, and disagreements between authors were adjudicated by the opinion of a third author. For the citation search, articles that cited the identified and included articles at the full text level were extracted, and title, abstract, and full text screening were repeated as described above.

Two authors collected study demographic information from included articles (Table S3), and one author extracted data for use in meta‐analyses. Graphical data were transcribed to numerical data using a validated digital ruler that has been previously reported to have ‘excellent’ intra‐rater reliability for graphical to numerical data conversion under our use (n = 5, intraclass correlation coefficient: 0.9752, 95% CI: [0.9413, 1.0000]) [32, 33]. We further validated independent measurements performed by two authors and detected ‘excellent’ inter‐rater reliability (n = 15, interclass correlation coefficient: 0.997, 95% CI: [0.990, 0.999]).

2.1.1. Reporting Quality, EV‐Reporting Transparency, and SYRCLE Risk of Bias assessments

Two authors independently assessed the reporting quality of the included studies using the full Animal Research: Reporting In Vivo Experiments (ARRIVE) Guidelines 2.0 [34]. The 21 categories were ranked as either “clearly insufficient” (0), “unclear if sufficient” (1), or “clearly sufficient” (2). The highest possible score of 42 indicates more transparent reporting, while a 0 indicates no transparency. Scores between the two authors were averaged. Spearman's correlation analysis between ARRIVE score and year of article publication was calculated via SPSS Statistics for Windows, version 31 (IBM Corp., Armonk, NY, USA).

Additionally, two independent authors used the Minimal Information for Studies of Extracellular Vesicles (MISEV) Guidelines to evaluate transparency in reporting of key EV isolation methods and characteristics [25, 35, 36]. Categories were treated as binary, with a 1 indicating that category was reported and a 0 indicating that category was not reported. Similar to the ARRIVE scores, we correlated the total number of reported items with year of publication using a Spearman correlation analysis in SPSS Statistics. In our supplementary table, we include guidelines from all three versions (2014, 2018, 2024), but in our figure and analyses, we only include guidelines from 2018, given that all papers were published after this date and had the opportunity to ensure compliance with the guidelines.

Lastly, two authors independently assessed the SYstematic Review Centre for Laboratory animal Experimentation (SYRCLE) risk of bias (ROB) in publications guidelines [37]. The ten categories were ranked as either “high risk” (0), “unclear risk” (1), or “low risk” (2). The highest possible score of 20 indicates the lowest risk of bias, while 0 indicates a relatively high risk of bias. Scores between the two authors were averaged. Spearman's correlation analysis between SYRCLE ROB score and year of article publication was calculated via SPSS Statistics. We note that while the validated SYRCLE uses scoring of high risk/unclear risk/low risk, the conversion of this to a numerical scale and ultimate use in a correlation analysis is not formally validated and is used here as an exploratory analysis.

2.2. Dose Response Curve and Visualization

In an attempt to examine the effect of dose on histological measure of tendon healing, we visualized the relationship between dose with the fixed effect size of histological outcome of tendon healing. Manuscripts reported EV doses using two different methods: protein in micrograms (µg) or particle number, normalized to animal weight and volume of injection. Therefore, we visualized two relationships: one including studies that reported EV dose using EV protein mass and another including studies that reported number of EV particles administered. We plotted inverse variance as the marker size to provide insight into study reliability. Only studies in which one dose of EVs was administered immediately after the injury were included in this visualization to reduce the number of confounding variables.

2.3. Meta‐Analyses

We performed meta‐analyses on histological scores of tendon healing, maximum load to failure, stiffness, Young's modulus, type I & III collagen protein and transcript expression, IL‐6 protein expression, and protein markers of senescence (p16, p21, senescence‐associated beta‐galactosidase (SA‐β‐GAL)). These outcomes were selected as they were the only ones with more than four independent studies with quantification. If a study reported multiple outcomes (e.g., a study reported histological score based on H&E and quantified load to failure), those outcomes were included in both analyses. We calculated the standardized mean differences (SMD) and pooled standard deviations (SD_pool) of these variables from included studies via the DerSimonian‐Laird method, which accounts for between‐study heterogeneity under a random‐effects framework and permits pooling across studies using different outcome measures and scoring systems [38]. We chose these specific methods because the outcomes assessed in our meta‐analyses include comparable outcomes with different measurement scales. We selected a random‐effects model using the DerSimonian‐Laird estimator to account for anticipated between‐study heterogeneity. For histological scoring, we oriented all scores such that a higher score indicates higher quality tendon healing, and a lower score indicates more pathology. For studies with multiple time points and multiple treatment dosages, the time point furthest out from injury and/or the highest dose were included. For studies with multiple treatment arms, only native EVs were utilized in our analyses (i.e., no engineered EVs). If studies had a native, manipulated, and a control group, only the native group was included in our analysis. Cochran's Q test was utilized to assess heterogeneity and homogeneity, and statistical significance was defined as an alpha level of 0.05. Meta‐analyses with random effects were also performed via SPSS Statistics for Windows, version 31 with no small‐sample correction added. We chose not to include a small‐sample correction, given that the sample sizes varied from small to moderate to large samples. Specifically, mean, sample size, and standard deviation from the original studies were input into SPSS to calculate the effect size, 95% confidence interval, and weight of each study within the meta‐analysis.

3. Results

3.1. Wide Variability in EV Origin, Characterization, and In Vivo Injury Models

Following database searches, we identified 344 unique studies related to EV exposure and tendon healing following an injury. After title, abstract, and full text screening, 42 studies were considered eligible for inclusion in our review. After a citation search, we identified four additional studies for inclusion, leading to 46 studies being included in our analyses (Figure 1) [39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84].

Figure 1.

Systematic review workflow. After title, abstract, and full text screening as well as a citation search, 46 articles were determined to be eligible for inclusion in this review. Reasons for exclusion at the full text level are reported in Table S2 and basic study information about included articles is listed in Table S3.

Eighteen studies (39%) isolated EVs from rat cells (Figure 2A), with bone marrow derived and adipose stem cells being the most frequent cell type (n = 10 & 11, 22 & 24% respectively, Figure 2B). Most studies utilized cells from passages 3–5 (n = 16–21, 35%–46%), while thirteen studies (27%) did not report what cell passage was utilized (Figure 2C). Thirty studies (65%) used EVs from a young cell source, while 13 studies (28%) did not report the age of the cells utilized (Figure 2D). Similarly, 23 studies (50%) utilized only male cells to generate their EVs, and 16 studies (35%) did not report sex (Figure 2E). CD63 (n = 24, 52%), CD9 (n = 26, 57%), and TSG101 (n = 22, 48%) were the most common markers utilized for EV characterization (Figure 2F). Notably, few studies reported markers to ensure purity of EV samples (e.g., ApoA1, ApoB), as recommended by the MISEV Guidelines [25]. Most studies included EVs that ranged from 100 to 300 nm in diameter (n = 28–45, 61%–98%) (Figure 2G).

Figure 2.

Extracellular vesicle (EV) and in vivo demographic information from included studies. Numbers over bars are # of studies (% of studies). (A) EV Species. (B) EV Source. Adipose stem/stromal cells (ASC). Bone marrow derived stem/stromal cells (BMSC). Dendritic cells (DC). Mesenchymal stem cells (MSCs) derived from induced pluripotent stem cells (iMSC). Macrophages (M). Platelets (P). Tendon stem/progenitor cells (TSPC). Umbilical cord stem cells (UCSC). Urine stem cells (USC). (C) Passage number of cells that produced EVs. (D) Age of cell that produced EVs. (E) Sex of cell that produced EVs. (F) Markers used to categorize EVs. (G) EV Size included in study. (H) Species of in vivo animal model. (I) Age of in vivo animal model. (J) Sex of in vivo animal model. (K) Tendon that was injured. (L) Injury model utilized. (M) EV delivery methodology. (N) Biomaterial used to supplement EVs.

Examining in vivo model selection, 34 studies (74%) used a rat model, 42 studies (91%) utilized young animals, and 41 studies (89%) used only males (Figure 2H–J). While there was a diversity of tendons selected for injury, Achilles tendon injuries were utilized in 24 studies (52%) (Figure 2K). In 29 studies (63%), tendons were injured via full transection followed by surgical repair, in six studies (13%) a partial transection and repair were performed, and 12 studies (26%) utilized a collagenase injection (Figure 2L). Thirty‐one studies (67%) delivered EVs a single time immediately after the injury (Figure 2M). Fourteen studies (30%) delivered EVs with a gelatin support matrix, four studies (9%) provided a collagen‐based support matrix, and 15 studies (33%) provided no biomaterial support (Figure 2N).

3.2. Exclusion Criteria, Blinding, Animal Housing Details, and Protocol Registration Were Poorly Reported

Next, we assessed reporting of included articles using the ARRIVE guidelines (Table S4). Study design, outcome measures, results, background, abstract, objectives, ethical statement, interpretation, data access, and conflicts of interest were adequately reported in the literature. Sample size, randomization, statistical methods, experimental animals, experimental procedures, and generalizability/translatability were unclear if adequately reported. Finally, inclusion/exclusion criteria, blinding, animal housing & husbandry, animal care & monitoring, and protocol registration were inadequately reported (Figure 3A). Examining the relationship between transparency of reporting and year of publication, we note that all included studies were published after 2018. In the infancy of this literature, there was no statistically significant relationship between transparency in reporting and time (Figure 3B).

Figure 3.

ARRIVE, MISEV, and SYRCLE ROB Reporting Guideline Scores. (A) ARRIVE guideline category assessment for included studies. (B) Relationship between year of publication and transparency in reporting, as assessed with the ARRIVE Guidelines. Table S4 includes individual score per paper. (C) 2018 MISEV guideline category assessment for included studies. (D) Relationship between year of publication and transparency in reporting, as assessed with the 2018 MISEV Guidelines. Table S5 includes individual score per paper as well as categories from the 2023 guidelines. (E) SYRCLE Risk of Bias category assessment for included studies. (F) Relationship between year of publication and risk of bias, as quantified by the SYRCLE criteria. Table S6 includes individual score per paper.

3.3. Several EV‐Specific Parameters Remain Poorly Reported in the Literature

Similarly, we also assessed the transparent reporting of EV characteristics using the MISEV guidelines. The 2018 MISEV Guidelines were used for analyses given all included manuscripts were published after 2018 (Table S5) [25, 35, 36]. When looking at the specific categories, we found most papers appropriately reported information on media composition, producing cell characteristics, culture temperature, gas concentration, temperature during concentrating, EV diameter distribution, and EV morphology. However, cell seeding density, molecular weight cut‐offs, tube type, settings during centrifugation, protein concentration in EVs, and confidence interval of EV size remain poorly reported (Figure 3C). We did not detect a statistically significant relationship between year of publication and adherence to the 2018 MISEV guidelines (Figure 3D).

3.4. The Risk of Bias in Publication in This Body of Literature Is Relatively High

When assessing adherence to SYRCLE ROB criteria, sequence generation was the only category consistently reported. Conversely, the remaining categories, including baseline characteristics, allocation concealment, random housing, blinding of intervention and outcome, random outcome assessment, incomplete outcome data, selective outcome reporting, and other sources of bias, had relatively high ROB (Figure 3E, Table S6). Moreover, we did not detect a statistically significant relationship between year of publication and risk of bias, as quantified by SYRCLE guidelines (Figure 3F).

3.5. EV Treatment Is Associated With Improved Histological Scores and Mechanical Properties of Tendon, Despite High Heterogeneity

Next, we performed meta‐analyses to assess the impact of EV treatment on histological score and mechanical properties. We found that EV treatment was associated with improved histological score when compared to tendons not treated with EVs (Figure 4A). Similarly, we found an association between improved mechanical properties and EV treatment. Specifically, the maximum load to failure (Figure 4B), tendon stiffness (Figure 4C), and Young's modulus (Figure 4D) increased in tendons treated with EVs relative to those that did not receive EV treatment. Interestingly, these differences were detected despite analyses exhibiting high heterogeneity (I² > 90%). Using the demographic information from Figure 2, we attempted to perform subgroup analyses to identify the source of this heterogeneity, but unfortunately, given the high variability and the relatively small number of included studies (e.g., Figure 2F), we were unable to identify a statistically significant source (Table S7). Specifically, we did not detect an effect of sex, species, age, size, cell type or passage from which EVs were obtained; EV markers used for characterization; in vivo species, sex, age, tendon, model, biomaterial, time after injury or EV administration tissues (Table S7). Notably, we were unable to assess for an effect of native versus primed cells because there were not enough outcomes in both groups (i.e., only one study included in the meta‐analyses used primed cells). Additionally, for all of these analyses, funnel plots do not depict symmetrical spread across the inverted pyramid, which suggests publication bias, small‐study effects, or other sources of between‐study variation may be contributing to high heterogeneity (Figure S1). Given the high heterogeneity amongst the histological outcomes, we next sought to examine the relationship between dosage of EVs and effect size of histological outcome. Unfortunately, we were not powered to evaluate this relationship statistically, but we did visualize these relationships in Figure S2A–S2B.

Figure 4.

Meta analyses assessing the effects of EVs on histological score, load to failure, stiffness, and Young's modulus. (A) Meta‐analysis on histological scoring. (B) Meta‐analysis on maximum failure load. (C) Meta‐analysis on tendon stiffness. (D) Meta‐analysis on Young's modulus. Square = effect size for individual studies, with line indicating 95% confidence interval and size indicating the weight of the analysis. Diamond = average effect size from all included data.

3.6. EV Treatment Is Associated With Increased Tendon Type I Collagen Protein Expression and Type III Collagen Transcript Expression

Lastly, we performed a series of meta‐analyses to identify mechanisms that may be contributing to improvements in tendon histological structure and mechanical properties. We found that EV treatment was associated with increased protein expression of type I collagen but not type III collagen relative to untreated controls (Figure 5A–5B). Interestingly, our meta‐analysis revealed the opposite for transcript expression; specifically, EV treatment was associated with increased type III but not type I collagen transcript expression relative to untreated controls (Figure 5C–5D). We also detected a significant association between decreased IL‐6 protein expression and markers of senescence in EV treatment relative to untreated controls (Figure S2C–S2E). Again, funnel plots demonstrated a lack of symmetry across the inverted pyramid, suggesting other sources of variability may be contributing to these findings (Figure S3). As before, given high heterogeneity, we attempted to disentangle potential co‐drivers of this relationship but did not detect any significant interactions (Table S7).

Figure 5.

Meta analyses assessing the effects of EVs on type I and III collagen protein and transcript expression. (A) Meta‐analysis on type I collagen protein expression. (B) Meta‐analysis on type III collagen protein expression. (C) Meta‐analysis on type I collagen transcript expression. (D) Meta‐analysis on type III collagen transcript expression. Square = effect size for individual studies, with line indicating 95% confidence interval and size indicating the weight of the analysis. Diamond = average effect size from all included data.

4. Discussion

The purpose of this study was to systematically review the current body of peer‐reviewed literature to better understand the impact of EVs on tendon healing in pre‐clinical animal models. We found that most studies utilized young male animals, though a diversity of injury locations and types, EV dosage schemes, and biomaterials were implemented. Several features of rigor and reproducibility, such as exclusion criteria and blinding, as well as EV reporting guidelines, such as confidence interval of EV size, remain poorly reported in this body of literature, and risk‐of‐bias in publication is relatively high in this body of literature due to inadequate reporting. Despite high heterogeneity, meta‐analyses revealed that EV treatment was associated with improved histological evaluation of tendon healing and mechanical properties in preclinical models, including increased load to failure, tendon stiffness, and Young's modulus. EV treatment also increased type I collagen protein and type III collagen transcript expression, while decreasing protein expression of IL‐6 and markers of senescence.

Several studies have reported that tendon regeneration is mediated by age [85, 86] and sex [87, 88] and that EV effects on tissues are impacted by age and sex of the origin cells [89, 90]. Yet, these two variables are sparsely considered in this specific body of literature. The interacting effects of age and sex are particularly important given that post‐menopausal women have higher rates of tendon injuries than pre‐menopausal women and age‐matched men [91, 92], with documented effects of estrogens on tendon mechanical properties [93, 94]. Therefore, we recommend future studies integrate age‐by sex paradigms into their study design.

Although there was high heterogeneity in most of our meta‐analyses, we could not discern the origin of this heterogeneity. One area of particular interest is how dosage and frequency of treatment may impact outcome variables. To our surprise, few studies administered multiple doses of EVs and because of this, we were not able to quantitatively evaluate how this impacted outcomes. However, administering multiple doses over time may increase treatment efficacy, as several studies in different organ systems have supported a multi‐dose strategy [95, 96, 97]. We also visualized the relationship between dosage and effect size of histological outcomes due to not being statistically powered to formally analyze this relation and did not see any obvious patterns. It is unclear whether this is a true null finding or is due to the small sample sizes available in our analysis or other sources of variability in study designs. Nonetheless, EV dosage and frequency are two important variables that should be prioritized for targeted investigation in future studies.

One mechanistic insight gained from this study is the impact of EVs on type I and III collagen expression following tendon injury. In healthy tendons, approximately 95% of collagen is type I collagen [98]. Type III collagen expression, which has lower tensile strength, increases in the early stages following injury and is ultimately replaced by type I collagen, though this process can take years [99]. Our meta‐analyses were unable to discern a time‐based relationship between injury status and collagen turnover, likely due to variability among studies. Furthermore, while two studies in our review examined both transcript and protein expression of both type I and III collagen [42, 51], samples were all collected at the same time point, thus we were unable to evaluate how collagen transcripts and proteins change relative to each other across time. Understanding the relationship between collagen transcript and protein expression timing along the regenerative cascade is a critical area of clarification for future research to correctly interpret EV‐mediated effects on tissue repair.

Another question raised from our analyses is: are increases in collagen expression contributing productively to the regenerative cascade? Given that EV administration was also associated with improved mechanical properties and histological signs of healing, it would appear that increasing collagens is beneficial, but future mechanistic studies should consider the impact of EVs on collagen hydroxylation and further assess structure using techniques such as second harmonic generation or electron microscopy to ensure the newly deposited collagen is positively contributing to tendon structure and function.

Although this review adds important context to the growing body of literature investigating the potential therapeutic role of EVs on tendon healing, it does have some important limitations. Our review was not prospectively registered, and we excluded non‐English, non‐peer reviewed studies. The generalizability of these findings is limited to the demographics included in the studies (i.e., young, male tendons). Although this methodology is well established and was validated by a second author, all of the data utilized in our meta‐analysis were only extracted by a single author. Almost all of our meta‐analyses have high heterogeneity, and there are likely other variables impacting outcome variables that we could not identify. Specifically, the null meta‐analyses are challenging to interpret since we cannot identify if the study variability is contributing to the lack of null results or if they are true null results. We acknowledge that Hedges' g and a small‐sample size correction can reduce small‐sample bias which we did not use in our meta‐analysis due to the variability in sample size (small to moderate to large) and because our analysis used uncorrected SMDs. There is a reporting bias in the literature, with 33% not reporting the sex and 27% not reporting the age of the cells used to generate the cells, which may ultimately bias results. Additionally, for the protein and transcript meta‐analyses, we could not directly control for housekeeping protein/gene usage, which may have impacted results and makes pooling molecular outcomes with differing normalization metrics challenging. Given funnel plots demonstrated asymmetry, null findings or lower effect sizes may have been detected in other non‐published studies, and the lack of inclusion of this data may have affected meta‐analysis skew. Lastly, while we focused on in vivo studies, mechanistic insights can be gained from in vitro studies as well, as they enable isolation of specific cellular interactions, signaling pathways, and EV‐mediated effects under defined experimental conditions that are difficult to dissect in vivo.

5. Conclusions

The principal findings from this review are that EV treatment may be beneficial to tendon healing in preclinical models, as measured by histological outcome, load to failure, tendon stiffness, and Young's modulus. Mechanisms that may be contributing to these improved outcomes include EVs increasing type I collagen protein expression and type III collagen transcript expression, while decreasing protein expression of IL‐6 and senescence markers. Most published studies utilized young, male Achilles tendons. Further inclusion of aged and female animals will increase the translational potential of this body of literature. There is a significant amount of variability in injury type, dosage schema, and biomaterial utilization, and future studies should consider systematically testing the impact of these variables to discern their impact on EV effectiveness. Exclusion criteria, blinding, animal housing details, protocol registration, cell seeding density, gas concentration, tube type, and confidence interval of EV size remain poorly reported, and specific attention to more transparent reporting of these outcome variables should be included in future studies. Ultimately, these findings support further study of EV impacts on tendon healing and provide insights into how EVs may be enhancing healing.

Author Contributions

All authors made substantial contributions in the following areas: (1) conception and design of the study, acquisition of data, analysis and interpretation of data, drafting of the article; (2) final approval of the article version to be submitted; and (3) agreement to be personally accountable for the author's own contributions and to ensure that questions related to the accuracy are appropriately investigated, resolved and the resolution documented in the literature. The specific contributions of the authors are as follows: Conceptualization: Allison C. Bean. Methodology: Gabrielle Gilmer. Formal analysis: Gabrielle Gilmer. Investigation: Gabrielle Gilmer, Kariman Shama, Liam Klingensmith, Jesse Zhang, Allison C. Bean. Resources: Allison C. Bean. Data curation: Gabrielle Gilmer, Kariman Shama, Liam Klingensmith, Jesse Zhang, Allison C. Bean. Visualization: Gabrielle Gilmer. Project administration: Gabrielle Gilmer, Allison C. Bean. Supervision: Allison C. Bean. Writing – original draft: Gabrielle Gilmer. Writing – review and editing: Gabrielle Gilmer, Kariman Shama, Liam Klingensmith, Jesse Zhang, Brittany Taylor Allison C. Bean.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: Funnel plots for meta‐analyses. (A) Histological scoring of tendon healing. (B) Load to failure. (C) Tendon stiffness. (D) Tendon Young's Modulus

Figure S2: The relationship between EV dose and histological outcome and meta‐analyses on IL‐6 and markers of senescence. (A) Relationship between EV dose and histological outcome, with EV dose quantified in µg. (B) Relationship between EV dose and histological outcome, with EV dose quantified in particles. (C) Meta‐analysis on IL‐6 protein expression. (D) Meta‐analysis on markers of senescence protein expression. (E) Funnel plot on IL‐6 protein expression. (F) Funnel plot of senescence protein expression

Figure S3: Funnel plots for meta‐analyses. Type I collagen (A) protein and (B) transcript. Type III collagen (C) protein and (D) transcript

Supporting File 1

Supporting File 2

Supporting File 3

Data Availability Statement

The data that supports the findings of this study are available in the Supporting material of this article.