What was reviewed
Madina Kurmanalina, Nader Tanideh, Amin Tamadon and colleagues, at Shiraz University of Medical Sciences and West Kazakhstan Marat Ospanov Medical University, ran a PRISMA systematic review and meta-analysis of in vivo extracellular vesicle (EV) therapy in regenerative endodontics, registered with PROSPERO (registration 420251107328) and published in Science Progress, volume 109, issue 1, in 2026 (open access, DOI 10.1177/00368504261433133). They searched Web of Science Core Collection and Scopus from inception to 15 July 2025, with reference screening adding 27 records to the 119 database records, and a search update on 31 December 2025 that added nothing. After duplicate removal and screening, 21 in vivo studies of exosomes, microvesicles, or apoptotic bodies in pulp and dentoalveolar regeneration models qualified for qualitative synthesis. Seven reported enough data (postintervention means, standard deviations, sample sizes) for the mineralization meta-analysis, and five for angiogenesis. Risk of bias was scored with the SYRCLE tool for animal studies, and pooling used a random-effects model with the Hartung-Knapp-Sidik-Jonkman adjustment, chosen because fewer than ten studies sit behind each pooled estimate.
What they found
EV-based treatment raised mineralization versus controls with a pooled standardized mean difference (SMD) of 6.43 (95 percent CI 3.13 to 9.73, p = .0001), with very high heterogeneity (I2 = 91 percent, Q = 78.71, p less than .0001). Angiogenesis outcomes pooled to an SMD of 7.89 (95 percent CI 3.94 to 11.85, p less than .0001) across five studies, with I2 = 82 percent (Q = 39.71). Subgroup splits by EV cell source were uneven in strength of evidence. For mineralization, DPSC-derived EVs gave an SMD of 4.82 (95 percent CI 2.41 to 7.22, I2 = 0 percent) and hESC-derived MSC EVs gave 6.71 (4.53 to 8.90, I2 = 0 percent), while SHED-derived EVs gave 7.75 with a confidence interval of -3.34 to 18.85 that crosses zero, non-stem dental pulp cell EVs gave a nonsignificant 0.48 (-0.47 to 1.42), and a single SCAP study gave 14.13 (9.20 to 19.07). For angiogenesis, DPSC EVs across three studies gave an SMD of 11.88 (6.81 to 16.95, I2 = 65 percent), the single SCAP study gave 11.82 (5.17 to 18.48), and SHED EVs gave a smaller but precise 2.85 (1.69 to 4.01, I2 = 0 percent). The authors describe this as the first meta-analysis restricted to in vivo EV studies in regenerative endodontics.
What it does not show
The effect sizes are not clinically readable. SMDs of 6 to 8 in tightly controlled animal experiments mostly reflect large between-group differences against small within-group variance, and the authors themselves state that the pooled numbers support the direction of benefit, not a magnitude a clinician could expect. The quantitative base is thin: seven studies behind the mineralization number, five behind angiogenesis, drawn from mice, rats, miniature pigs, and dogs, plus one study that also used a human premolar fragment model, with follow-ups from 3 days to 6 months. Bias is the weak point, not a caveat. On the authors’ own SYRCLE counts, all 21 studies were rated unclear overall, only 5 reported enough on randomization for a low selection-bias rating, performance bias was rated low in all 21 (standardized housing), and none described blinding of outcome assessors, so detection bias is unclear across the board, which matters most exactly where outcomes are semiquantitative histology or immunostaining. Dosing was reported inconsistently (micrograms of protein, particle counts, or concentration per volume) with no normalization to defect size or body weight, and the included models range from direct pulp capping to Matrigel plug assays and subcutaneous implants, which is where the 82 to 91 percent heterogeneity comes from. Nothing here is human trial evidence, and the paper includes no publication-bias assessment (funnel analyses are unreliable below ten studies, which this review notes as the reason for the HKSJ adjustment).
Where it sits in the field
For the pulp and dentin repair program, this is the cell-free wing of pulp regeneration getting its first dedicated quantitative synthesis, and the direction matches the route’s individual primary studies, from SCAP exosomes driving dentin-pulp complex regeneration to apoptotic vesicles from dental pulp stem cells building vascularized pulp-like tissue in mouse and dog injury models. Two signals are useful for the program tracker. First, the cell-source ranking in the abstract (dental stem cell EVs strongest) is firmer for DPSC EVs than the headline suggests: the SHED mineralization interval crosses zero and rests on few studies, and non-stem pulp cell EVs show no significant effect, so “dental stem cell-derived” is doing real work in that claim. Second, the review names the translation blockers in plain terms: no dose normalization, scaffold-dependent delivery of unproven clinical handling, uncharacterized biodistribution and off-target effects, and no GMP-grade production, batch consistency, or potency assays yet. Those are the same manufacturing gaps the cell-free therapies field assessment tracks for EV products across this route. The current field assessment stands at /field/.
Where we differ from the coverage
We found no press or popular coverage of this paper to differ from. Against the abstract, we would not lead with “significantly increased mineralization and angiogenesis” without the heterogeneity figure in the same sentence: a 91 percent I2 means the pooled SMD of 6.43 is an average of studies that disagree with each other enormously, and the subgroup the text singles out as stronger for mineralization, SHED-derived EVs, has a confidence interval that includes no effect. We also read “risk of bias was predominantly unclear” more sharply than the abstract states it: on the paper’s own results, all 21 studies were rated unclear overall and none reported blinding of outcome assessors. For semiquantitative histology endpoints, that is not a footnote; it is the main reason to treat every point estimate above as provisional.
Provenance: source read in full (open-access full text via Europe PMC, PMC13039592); every number above as reported there. Selection and review procedure: /method/.