Non-Bone-Derived Extracellular Vesicles in Osteoporosis: Evidence Mapping and Nanomedicine Translation Authors: Jinyang An, Xinsai Li, Ying Yang, Yangyang Zhang, Shuyun Li, Mingming Jin, Haihong Lv Published: 2026 ABSTRACT BackgroundOsteoporosis is influenced by signals from multiple tissues and body systems. Extracellular vesicles (EVs) and EV-like particles may contribute to bone remodeling, serve as biomarkers, or act as therapeutic delivery platforms, but these distinct research settings are often conflated and their evidential and translational maturity remains unclear.ObjectiveTo map research on non-bone-derived EVs and EV-like particles in osteoporosis-related bone remodeling using a purpose-based framework and to assess source and cargo attribution, evidence strength, and translational stage.MethodsPubMed, Web of Science Core Collection, and EMBASE were searched from inception to July 15, 2026, supplemented by citation searching. Eligible original studies evaluated bone-related cellular or remodeling outcomes, clinically relevant bone phenotypes, biomarker associations, or therapeutic and delivery effects. Studies were classified as mechanism-, biomarker-, or intervention/delivery-focused, graded using an operational A-C evidence framework, and stratified from T0 to T3.ResultsWe included 114 studies: 39 mechanism-focused, 22 biomarker-focused, and 53 intervention/delivery-focused. Most evidence remained preclinical, with 24 T0 and 76 T1 studies; 14 studies reached participant-level T2 evidence. Although 55 studies met level C criteria, EV dependence, cargo attribution, and endogenous source-to-bone attribution represented distinct and non-interchangeable dimensions, and strict in vivo source-to-bone attribution remained uncommon. In the bibliographic databases searched, no eligible published clinical or industrial evidence at T3 was identified. External validation of human biomarkers, biodistribution assessment, standardized potency assays, and product-level quality control were also uncommon.ConclusionNon-bone-derived EVs and EV-like particles should be interpreted as a purpose-stratified evidence landscape rather than as a single biological mechanism. Current findings provide an evidence-organizing framework, not clinically ready diagnostic or therapeutic platforms. Priorities include stronger source tracing and cargo attribution, longitudinal human validation, quantitative biodistribution and potency assessment, and GMP-compatible product development. Introduction Osteoporosis: From Local Imbalance in Bone Metabolism to Extraosseous Multi-Axis Regulation Osteoporosis (OP) is a systemic skeletal disorder characterized by reduced bone mass, deterioration of bone microarchitecture, decreased bone strength, and an increased risk of fracture, with fragility fractures representing one of its most important clinical outcomes.1 In clinical practice, OP risk is primarily assessed on the basis of bone mineral density (BMD) measured by dual-energy X-ray absorptiometry (DXA), a history of previous fractures, the Fracture Risk Assessment Tool (FRAX), and bone turnover markers. However, even among individuals with similar BMD and conventional clinical risk factors, the rate of bone loss, bone microarchitecture, bone turnover status, and fracture risk may still differ.2,3 Therefore, novel biomarkers should not only demonstrate an association with OP but should also be evaluated for their ability to reflect disease progression or treatment response and to provide additional clinical information beyond that offered by DXA, FRAX, and existing bone turnover markers.4,5 For many years, OP has been interpreted primarily within an endocrine–metabolic framework, in which estrogen deficiency, disturbances in calcium–phosphate metabolism, and imbalanced bone turnover have been regarded as important drivers of the disease.6 Advances in osteoimmunology have further demonstrated that osteoblasts (OBs), osteoclasts (OCs), osteocytes, bone marrow mesenchymal stem cells (BMSCs), and immune cells continuously interact within a shared bone marrow niche, collectively maintaining the dynamic balance between bone formation and bone resorption.7–10 Under osteoporotic conditions, chronic low-grade inflammation, alterations in immune-cell profiles, and persistent activation of inflammatory signaling can promote bone resorption, suppress bone formation, and impair bone microarchitecture.11–14 OP and low bone mass also commonly coexist with conditions such as sarcopenia, vascular dysfunction, gut microbiota dysbiosis, and abnormal fat distribution.15–18 Although these comorbid associations do not, in themselves, demonstrate that extraosseous abnormalities directly cause bone loss, basic research suggests that muscle, adipose tissue, the vasculature, the immune system, the microbiota, and other distant organs can influence bone-cell function and bone-remodeling processes through mechanical, inflammatory, metabolic, and neuroendocrine pathways.9,19–28 Accordingly, OP is increasingly understood as a pathological process jointly shaped by an imbalance in local bone remodeling and multisystem signaling inputs originating outside the skeleton. Extracellular Vesicles as Nanoscale Mediators of Inter-Tissue Communication When OP is considered within a multi-organ and multisystem regulatory framework, an important question arises: how are changes in the states of extraosseous tissues transmitted to bone and subsequently translated into recognizable abnormalities in bone remodeling? Cytokines, chemokines, hormones, and soluble metabolites participate in communication between organs and bone; however, a single soluble factor is unlikely to simultaneously reflect the state of the donor cell, the composition of complex cargoes, and context-dependent effects.29,30 Extracellular vesicles (EVs) are nanoscale membrane-bound particles released by cells and can carry a wide range of bioactive components, including proteins, lipids, mRNAs, miRNAs, and other non-coding RNAs.31 Their membrane-surface characteristics and cargo composition can retain information about the state of the donor cell and simultaneously influence multiple cell types involved in bone remodeling.26,27,32–38 Therefore, EVs have the mechanistic potential to connect the states of extraosseous tissues with bone-remodeling responses. In addition, EVs can be detected in the circulation and taken up by bone-related cells, while some studies involving exogenous EV administration have reported improvements in bone-related phenotypes. These findings support, respectively, their in vivo accessibility, their capacity for uptake by recipient cells, and their potential for therapeutic intervention.39–42 Accordingly, EVs may be regarded as candidate nanoscale mediators linking changes in extraosseous tissue states to bone-remodeling responses. Current Status, Limitations, and Controversies in OP-Related EV Research In recent years, OP-related EV research has gradually shifted from conceptual discussion to empirical investigation.43,44 Research has expanded from early mechanistic studies of the regulation of OB, OC, and BMSC fate45–47 to circulating EV biomarkers, interventions using native cell-derived EVs, and engineered cargo delivery.44,48,49 Current OP-related EV research encompasses three distinct research contexts. The first evaluates whether EVs naturally produced within the host participate in OP-related bone remodeling and primarily addresses questions of disease mechanism.50 The second examines the relationships between the abundance, phenotypes, or cargoes of EVs in body fluids, such as blood and urine, and BMD, bone turnover, fracture risk, or disease stratification, and primarily addresses biomarker-related questions.24,48,51 The third uses isolated, cultured, naturally obtained, or engineered EVs and EV-like particles as therapeutic agents or cargo-delivery platforms and primarily addresses questions of intervention and delivery.44,49 These three research contexts, summarized in Figure 1, differ in the conditions under which the particles are generated, the certainty of source identification, evidentiary requirements, and translational objectives.Figure 1Conceptual framework for non-bone-derived extracellular vesicles in osteoporosis research. (A) Mechanistic studies investigate extracellular vesicle-mediated communication between non-bone-derived tissues or cells, including immune cells, muscle tissue, adipose tissue, vascular cells, and microbial sources, and the bone microenvironment, with the aims of elucidating the mechanisms underlying bone remodeling and identifying potential therapeutic targets. (B) Biomarker studies evaluate heterogeneous extracellular vesicle populations in blood and urine for the prediction, diagnosis, and monitoring of osteoporosis, with the translational goal of developing minimally invasive or non-invasive diagnostic and prognostic tools. (C) Therapeutic intervention and delivery studies examine natural extracellular vesicles derived from adult tissues or cells, perinatal tissues, and pluripotent stem cells; engineered extracellular vesicles carrying targeting ligands, peptides, small-molecule drugs, or miRNA/siRNA; and plant- or animal-derived extracellular vesicle-like particles, with the goal of advancing clinical evaluation and therapeutic application.Diagram: extracellular vesicle research in osteoporosis - mechanisms, biomarkers, therapies.The diagram has three panels on extracellular vesicle (EV) research in osteoporosis. Panel A shows non-bone tissues like immune cells, muscle, adipose, vascular cells and microbes sending EV signals to the bone microenvironment, impacting BMSCs, OBs, OCs and osteocytes. The goal is to understand EV-bone communication for therapeutic targets. Panel B focuses on biomarker screening from blood and urine, targeting diverse EV populations to find biomarkers for predicting, diagnosing, or monitoring osteoporosis, aiming for non-invasive diagnostics. Panel C discusses natural EVs from adult and perinatal tissues, engineering strategies with targeting ligands, peptides, miRNA or siRNA and small-molecule drugs, plus EV-like particles from plants and animals. The aim is to develop EV-based therapies for clinical evaluation and application.Abbreviations: BMSCs, bone marrow mesenchymal stem cells; EVs, extracellular vesicles; miRNA, microRNA; OBs, osteoblasts; OCs, osteoclasts; siRNA, small interfering RNA. Conceptual framework for non-bone-derived extracellular vesicles in osteoporosis research. (A) Mechanistic studies investigate extracellular vesicle-mediated communication between non-bone-derived tissues or cells, including immune cells, muscle tissue, adipose tissue, vascular cells, and microbial sources, and the bone microenvironment, with the aims of elucidating the mechanisms underlying bone remodeling and identifying potential therapeutic targets. (B) Biomarker studies evaluate heterogeneous extracellular vesicle populations in blood and urine for the prediction, diagnosis, and monitoring of osteoporosis, with the translational goal of developing minimally invasive or non-invasive diagnostic and prognostic tools. (C) Therapeutic intervention and delivery studies examine natural extracellular vesicles derived from adult tissues or cells, perinatal tissues, and pluripotent stem cells; engineered extracellular vesicles carrying targeting ligands, peptides, small-molecule drugs, or miRNA/siRNA; and plant- or animal-derived extracellular vesicle-like particles, with the goal of advancing clinical evaluation and therapeutic application. Previous reviews have generally organized relevant studies according to EV source, cargo molecules, or bone-remodeling signaling pathways, but have rarely distinguished systematically among the different research purposes described above or simultaneously evaluated source attribution, cargo attribution, evidence strength, and translational stage.26,52 Consequently, the disease mechanisms, biomarker value, and therapeutic or delivery potential supported by different studies may be interpreted as though they belong to the same level of evidence. The reproducibility of EV research is also affected by factors such as particle nomenclature, sample preprocessing, isolation and purification, and functional attribution. The 2023 Minimal Information for Studies of Extracellular Vesicles (MISEV2023) guidelines recommend that particles be named cautiously on the basis of the available experimental evidence and that sample processing, isolation and characterization, and functional evaluation be reported in sufficient detail.39 OP-related EV research has also expanded from the administration of native EVs to strategies involving donor-cell engineering, therapeutic RNA loading, and bone-targeting surface modifications, indicating a trend toward the evolution of EVs from biological information carriers into engineerable nanoscale delivery platforms.44,53 Early clinical translational efforts have also emerged in bone- and joint-related fields; however, OP-specific human studies and evidence of clinical application remain limited. Accordingly, this review organizes the evidence according to three primary research purposes—mechanism, biomarker, and intervention/delivery—and compares the strength of attribution and translational maturity across studies. Methods Study Design, Registration, and Research Questions This study used a scoping review design to systematically map the distribution of evidence on non-bone-derived EVs and EV-like particles in OP-related bone remodeling and to compare their evidence base, strength of attribution, and translational stage when investigated as endogenous mechanistic mediators, biomarkers, or intervention/delivery platforms. The study design followed the methodological framework for scoping reviews proposed by Arksey and O’Malley54 and was further guided by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews (PRISMA-ScR) in study identification, literature screening, data extraction, and results reporting.55 The following research questions were specified a priori: 1) How is the evidence on non-bone-derived EVs and EV-like particles distributed across mechanism-, biomarker-, and intervention/delivery-focused studies? 2) To what extent have individual studies established evidence chains linking donor source or donor state, EVs or EV-like particles, key cargoes, recipient cells, and bone-related phenotypes? 3) What sample types, OP definitions, and clinical endpoints have been used in existing human studies, and have these studies performed adjustment for confounding factors, external validation, or assessments of incremental clinical value? 4) What are the evidence strength and translational stage of the existing studies, and what are the principal evidence gaps? This scoping review was registered with the Open Science Framework (OSF) Registries on May 6, 2026. The registration materials documented the review questions, eligibility criteria, search strategy, data extraction framework, evidence-strength grading, translational-stage stratification, and evidence-synthesis methods. At the time of manuscript submission, the OSF registration remained under embargo, and a view-only access link was provided to the editors and reviewers: [https://osf.io/m85fj/overview?view_only=bec08bb3580d40daa5e49bc6a492c4b5]. Review Scope, Source Definition, and Particle Nomenclature Operational Definition of “Non-Bone-Derived EVs” This review focuses on the effects of non-bone-derived EVs and EV-like particles as inputs into OP-related bone remodeling. The term “non-bone-derived EVs” is used as an operational umbrella term that excludes EVs derived from cells directly involved in bone remodeling, including OBs, osteocytes, OCs, and BMSCs.39,56 This exclusion does not diminish the importance of bone-derived EVs; rather, it is intended to distinguish local intercellular communication within bone tissue from the effects of extraosseous signals on bone remodeling. By contrast, “extraosseous EVs” is a more anatomically restrictive term referring to EVs whose donor cells or tissues are located outside the bone and bone marrow, such as EVs derived from muscle or adipose tissue. “Non-osteogenic-lineage-derived EVs” merely describes whether the donor cells belong to the osteogenic lineage and is not used synonymously with either of the preceding terms. Although OCs originate from the monocyte–macrophage lineage rather than the osteogenic lineage, they remain core effector cells in bone remodeling.57,58 Therefore, OC-derived EVs are classified as bone-derived EVs and excluded from this review. Body Fluid Samples and Source Classification In this review, serum, plasma, urine, and other body fluids are regarded primarily as sampling media for EVs rather than as clearly defined donor sources equivalent to muscle, adipose tissue, or the vasculature, because body fluid EVs generally comprise heterogeneous mixtures of particles released by multiple cell and tissue types.59,60 Studies that measured only the abundance, phenotypes, or cargoes of body fluid EVs and analyzed their associations with BMD, bone turnover, or fracture risk were classified as biomarker evidence. When source attribution was established using methods such as donor-cell-specific markers or immunocapture, the study was further classified under the corresponding cellular or organ source. Particle Nomenclature In accordance with MISEV2023, “EV” was used as an umbrella term when the specific biogenesis pathway of the particles had not been sufficiently demonstrated.39 However, when original studies explicitly used terms such as exosomes, small extracellular vesicles (sEVs), microvesicles, bacterial outer membrane vesicles (OMVs), or plant-derived exosome-like nanoparticles, the original terminology was retained to avoid redefining the study objects beyond the scope of the original evidence. Microbial-derived vesicles, plant-derived exosome-like nanoparticles, animal-derived EVs, and other exogenous vesicle-like particles were included on the basis of their vesicle-like structures and functional properties related to the regulation of bone remodeling or cargo delivery. Their inclusion does not imply that these particles are equivalent to eukaryotic cell-derived EVs in terms of biogenesis, molecular composition, quality assessment, or regulatory classification.39,61–63 Literature Search Strategy The initial search was conducted on February 2, 2026, and the searches of PubMed, Web of Science Core Collection, and EMBASE were updated and rerun on July 15, 2026. Each database was searched from inception to the date of the final updated search, with no lower limit on publication year. The search strategies combined controlled vocabulary and free-text terms related to OP, EVs, exosomes, and relevant vesicle-like particles and were adapted to the search syntax of each database. In addition, the reference lists of the included studies and relevant reviews were screened to identify studies that may have been missed by the database searches. Gray literature, conference abstracts, preprints, institutional reports, and clinical trial registries were not systematically searched. Accordingly, the assessment of clinical or industrial translational stages was restricted to published studies retrieved from the bibliographic databases that met the prespecified eligibility criteria. The complete search strategies for all databases are provided in Supplementary Methods S1. Study Inclusion and Exclusion Criteria Studies were included if they met all of the following criteria: 1)They were published original in vitro, ex vivo, in vivo animal, or participant-level human studies; 2) The study context involved OP or OP-related bone loss, or the study directly evaluated bone-remodeling processes closely related to the development or progression of OP; 3) The exposure, intervention, or primary object of interest was a non-bone-derived EV or EV-like particle that met the operational definition used in this review; 4) The study evaluated bone-related cellular functions, bone-remodeling processes, OP-related bone phenotypes, biomarker associations, or therapeutic and delivery effects; 5) At least one item of information could be extracted regarding the donor source or state, EV or EV-like particle, key cargo, recipient cell, or bone-related phenotype. 1)They were published original in vitro, ex vivo, in vivo animal, or participant-level human studies; 2) The study context involved OP or OP-related bone loss, or the study directly evaluated bone-remodeling processes closely related to the development or progression of OP; 3) The exposure, intervention, or primary object of interest was a non-bone-derived EV or EV-like particle that met the operational definition used in this review; 4) The study evaluated bone-related cellular functions, bone-remodeling processes, OP-related bone phenotypes, biomarker associations, or therapeutic and delivery effects; 5) At least one item of information could be extracted regarding the donor source or state, EV or EV-like particle, key cargo, recipient cell, or bone-related phenotype. They were published original in vitro, ex vivo, in vivo animal, or participant-level human studies; 2) The study context involved OP or OP-related bone loss, or the study directly evaluated bone-remodeling processes closely related to the development or progression of OP; 3) The exposure, intervention, or primary object of interest was a non-bone-derived EV or EV-like particle that met the operational definition used in this review; 4) The study evaluated bone-related cellular functions, bone-remodeling processes, OP-related bone phenotypes, biomarker associations, or therapeutic and delivery effects; 5) At least one item of information could be extracted regarding the donor source or state, EV or EV-like particle, key cargo, recipient cell, or bone-related phenotype. Studies were excluded if they met any of the following criteria: 1)They were non-original publications; 2) They neither included an OP or related bone-loss context nor evaluated bone-remodeling outcomes relevant to the development or progression of OP; 3) The EVs were derived from core bone-remodeling cells, including OBs, osteocytes, OCs, or BMSCs; 4) The study focused primarily on general fracture healing, bone-defect repair, tissue-engineered bone regeneration, periodontal bone regeneration, or other bone disorders unrelated to OP; 5) The study did not evaluate bone-related cells, bone remodeling, OP-related phenotypes, biomarkers, or intervention/delivery outcomes; 6) Information on particle preparation, isolation, basic characterization, or source was seriously insufficient, making it impossible to determine reasonably whether the study object or observed effects were related to EVs or EV-like particles; 7) The study was a duplicate publication, had been retracted, or had no accessible full text. 1)They were non-original publications; 2) They neither included an OP or related bone-loss context nor evaluated bone-remodeling outcomes relevant to the development or progression of OP; 3) The EVs were derived from core bone-remodeling cells, including OBs, osteocytes, OCs, or BMSCs; 4) The study focused primarily on general fracture healing, bone-defect repair, tissue-engineered bone regeneration, periodontal bone regeneration, or other bone disorders unrelated to OP; 5) The study did not evaluate bone-related cells, bone remodeling, OP-related phenotypes, biomarkers, or intervention/delivery outcomes; 6) Information on particle preparation, isolation, basic characterization, or source was seriously insufficient, making it impossible to determine reasonably whether the study object or observed effects were related to EVs or EV-like particles; 7) The study was a duplicate publication, had been retracted, or had no accessible full text. They were non-original publications; 2) They neither included an OP or related bone-loss context nor evaluated bone-remodeling outcomes relevant to the development or progression of OP; 3) The EVs were derived from core bone-remodeling cells, including OBs, osteocytes, OCs, or BMSCs; 4) The study focused primarily on general fracture healing, bone-defect repair, tissue-engineered bone regeneration, periodontal bone regeneration, or other bone disorders unrelated to OP; 5) The study did not evaluate bone-related cells, bone remodeling, OP-related phenotypes, biomarkers, or intervention/delivery outcomes; 6) Information on particle preparation, isolation, basic characterization, or source was seriously insufficient, making it impossible to determine reasonably whether the study object or observed effects were related to EVs or EV-like particles; 7) The study was a duplicate publication, had been retracted, or had no accessible full text. Study Screening and Selection Before formal screening and assessment, two reviewers conducted a calibration exercise using a prespecified set of representative studies. The calibration set encompassed different primary research purposes, evidence grades, and translational stages. The two reviewers first independently performed pilot assessments according to the operational criteria presented in Table 1. They then compared their assessments and discussed any discrepancies to align their interpretation of the criteria for assigning A–C evidence grades and T0–T3 translational stages. The calibration exercise was conducted solely to clarify the application of the prespecified criteria; the grading principles were not modified on the basis of the study findings. Formal study screening commenced after calibration.Table 1Operational Definitions and Criteria for Evidence Strength Grading and Translational Stage ClassificationClassification DimensionLevelDefinitionOperational CriteriaEvidence strengthAPhenomenological or associative evidenceReports EV-related differences, associations, uptake, or bone-related phenotypic changes without experimental testing of the underlying mechanism, EV dependence, cargo attribution, or endogenous source contribution.Evidence strengthBMechanistic-support evidenceDemonstrates an EV-associated functional effect and provides at least one experimentally supported mechanistic link, such as candidate cargo or pathway changes, target validation without direct cargo perturbation, downstream pathway perturbation, or biodistribution–function concordance. Direct EV, cargo, or endogenous-source attribution is not established.Evidence strengthC-EVEV-dependence evidenceShows that the principal effect is altered by EV-release inhibition, EV depletion, uptake blockade, or an equivalent EV-directed intervention. Rescue with isolated EVs provides additional support but is not mandatory.Evidence strengthC-CargoCargo-attribution evidenceShows that perturbation of a defined EV cargo by knockdown, inhibition, overexpression, selective loading, depletion, or reconstitution alters the principal EV effect. Differential expression, target prediction, or pathway correlation alone is insufficient.Evidence strengthC-EndoEndogenous source-to-bone attribution evidenceDemonstrates in vivo that source-restricted manipulation of EV production or release from a defined cell or tissue alters a corresponding bone-related phenotype. Source-specific tracing or rescue may provide additional support. Exogenous EV administration or tracing alone is insufficient.Translational stageT0In vitro or ex vivo evidenceThe principal conclusion is supported only by cell-based, organoid, isolated-tissue, or ex vivo experiments. Human-derived materials alone do not qualify as T2 evidence.Translational stageT1Animal in vivo evidenceThe principal conclusion is supported by animal in vivo evidence, including disease-model validation, EV administration, genetic manipulation, biodistribution or bone-homing assessment, safety evaluation, or bone-phenotype analysis. Human-derived EVs tested in animals remain T1.Translational stageT2Participant-level human evidenceDirectly evaluates EV abundance, phenotype, or cargo in clinically characterized participants in relation to bone mineral density, bone turnover, fracture, disease classification, diagnosis, prognosis, or treatment response. Functional studies using participant-derived EVs qualify only when linked to the corresponding participant-level clinical phenotype.Translational stageT3Clinical or substantive industrial translational evidenceIncludes prospective clinical intervention studies, established clinical applications, or documented product development beyond laboratory-scale proof-of-concept, such as scalable GMP-compatible manufacturing, validated quality-control and potency systems, formal product-level safety evaluation, product registration, or regulatory advancement. Animal efficacy or laboratory engineering alone is insufficient.Notes: Evidence strength and translational stage were evaluated independently. Evidence strength increases from A to C, whereas translational stage advances from T0 to T3. C-EV, C-Cargo, and C-Endo represent distinct attribution domains and may be assigned individually or in combination; fulfillment of one domain does not imply fulfillment of the others. Each study was assigned an overall evidence-strength grade of A, B, or C, together with any applicable C-domain labels, and the highest translational stage directly supporting its main research question and principal conclusion. This framework was developed for evidence mapping and narrative synthesis and should not be interpreted as a formal risk-of-bias assessment or a general evidence-quality grading system.Abbreviations: EV, extracellular vesicle; GMP, good manufacturing practice. Table 1Operational Definitions and Criteria for Evidence Strength Grading and Translational Stage ClassificationClassification DimensionLevelDefinitionOperational CriteriaEvidence strengthAPhenomenological or associative evidenceReports EV-related differences, associations, uptake, or bone-related phenotypic changes without experimental testing of the underlying mechanism, EV dependence, cargo attribution, or endogenous source contribution.Evidence strengthBMechanistic-support evidenceDemonstrates an EV-associated functional effect and provides at least one experimentally supported mechanistic link, such as candidate cargo or pathway changes, target validation without direct cargo perturbation, downstream pathway perturbation, or biodistribution–function concordance. Direct EV, cargo, or endogenous-source attribution is not established.Evidence strengthC-EVEV-dependence evidenceShows that the principal effect is altered by EV-release inhibition, EV depletion, uptake blockade, or an equivalent EV-directed intervention. Rescue with isolated EVs provides additional support but is not mandatory.Evidence strengthC-CargoCargo-attribution evidenceShows that perturbation of a defined EV cargo by knockdown, inhibition, overexpression, selective loading, depletion, or reconstitution alters the principal EV effect. Differential expression, target prediction, or pathway correlation alone is insufficient.Evidence strengthC-EndoEndogenous source-to-bone attribution evidenceDemonstrates in vivo that source-restricted manipulation of EV production or release from a defined cell or tissue alters a corresponding bone-related phenotype. Source-specific tracing or rescue may provide additional support. Exogenous EV administration or tracing alone is insufficient.Translational stageT0In vitro or ex vivo evidenceThe principal conclusion is supported only by cell-based, organoid, isolated-tissue, or ex vivo experiments. Human-derived materials alone do not qualify as T2 evidence.Translational stageT1Animal in vivo evidenceThe principal conclusion is supported by animal in vivo evidence, including disease-model validation, EV administration, genetic manipulation, biodistribution or bone-homing assessment, safety evaluation, or bone-phenotype analysis. Human-derived EVs tested in animals remain T1.Translational stageT2Participant-level human evidenceDirectly evaluates EV abundance, phenotype, or cargo in clinically characterized participants in relation to bone mineral density, bone turnover, fracture, disease classification, diagnosis, prognosis, or treatment response. Functional studies using participant-derived EVs qualify only when linked to the corresponding participant-level clinical phenotype.Translational stageT3Clinical or substantive industrial translational evidenceIncludes prospective clinical intervention studies, established clinical applications, or documented product development beyond laboratory-scale proof-of-concept, such as scalable GMP-compatible manufacturing, validated quality-control and potency systems, formal product-level safety evaluation, product registration, or regulatory advancement. Animal efficacy or laboratory engineering alone is insufficient.Notes: Evidence strength and translational stage were evaluated independently. Evidence strength increases from A to C, whereas translational stage advances from T0 to T3. C-EV, C-Cargo, and C-Endo represent distinct attribution domains and may be assigned individually or in combination; fulfillment of one domain does not imply fulfillment of the others. Each study was assigned an overall evidence-strength grade of A, B, or C, together with any applicable C-domain labels, and the highest translational stage directly supporting its main research question and principal conclusion. This framework was developed for evidence mapping and narrative synthesis and should not be interpreted as a formal risk-of-bias assessment or a general evidence-quality grading system.Abbreviations: EV, extracellular vesicle; GMP, good manufacturing practice. Operational Definitions and Criteria for Evidence Strength Grading and Translational Stage Classification Notes: Evidence strength and translational stage were evaluated independently. Evidence strength increases from A to C, whereas translational stage advances from T0 to T3. C-EV, C-Cargo, and C-Endo represent distinct attribution domains and may be assigned individually or in combination; fulfillment of one domain does not imply fulfillment of the others. Each study was assigned an overall evidence-strength grade of A, B, or C, together with any applicable C-domain labels, and the highest translational stage directly supporting its main research question and principal conclusion. This framework was developed for evidence mapping and narrative synthesis and should not be interpreted as a formal risk-of-bias assessment or a general evidence-quality grading system. Abbreviations: EV, extracellular vesicle; GMP, good manufacturing practice. All records retrieved from the database searches were imported into Rayyan for management. Before formal screening, duplicates were identified using the duplicate-detection function in Rayyan and were manually verified and removed on the basis of titles, authors, publication years, journals, DOIs, and other bibliographic information. Standalone errata or correction notices and retraction notices identified through the database searches were not considered original research reports and were therefore removed before title and abstract screening. After deduplication and preprocessing, two reviewers independently screened the titles and abstracts according to the prespecified inclusion and exclusion criteria. Records were excluded at this stage if they were clearly non-original publications; were unrelated to EVs or EV-like particles; involved study objects or target tissues outside the scope of the review; did not address OP, OP-related bone loss, or bone-remodeling processes closely related to the development or progression of OP; or involved EV sources that did not meet the operational definition used in this review. Records that passed title and abstract screening underwent full-text eligibility assessment. The two reviewers independently evaluated whether the study context and outcomes were relevant to OP-related bone remodeling, whether the particle source met the operational definition of non-bone-derived EVs and EV-like particles used in this review, and whether the information on particle preparation, characterization, functional evaluation, and source attribution met the minimum requirements for inclusion in the analysis. The initial independent assessments of both reviewers were retained before consensus discussions. Because the A–C evidence grades and T0–T3 translational stages were ordinal categories, inter-rater agreement was assessed using linearly weighted Cohen’s κ. Any disagreements were discussed by the two reviewers with reference to the full texts of the original studies and the prespecified criteria. Disagreements that remained unresolved were adjudicated by the corresponding author. The final consensus assessments were used for evidence-mapping analyses and were reported in the supplementary tables. The processes of study identification, title and abstract screening, full-text eligibility assessment, and final inclusion were presented in a PRISMA-ScR flow diagram (Figure 2).Figure 2PRISMA-ScR flow diagram of the study-selection process. The diagram summarizes study identification through database and citation searching, removal of duplicate and ineligible records, title and abstract screening, full-text eligibility assessment, and final inclusion. A total of 99 studies were included through database searching and 15 through citation searching, resulting in 114 studies included in the review.A flowchart of study identification and screening process, showing steps from identification to inclusion.The flowchart illustrates the study identification and screening process. It begins with identification via databases and registers, with records identified from databases totaling 3952, including PubMed, Web of Science and EMBASE. Records removed before screening include duplicates, ineligible records and corrections. Records screened number 2606, with exclusions for non-original studies, unrelated topics and ineligible subjects. Reports sought for retrieval total 638, with none unretrieved. Reports assessed for eligibility also total 638, with exclusions for ineligible subjects, models and sources. Identification via other methods includes records from websites, organizations and citation searching, totaling 15. Reports sought for retrieval and assessed for eligibility both number 15, with no exclusions. The process concludes with 114 studies included in the review.Abbreviations: EVs, extracellular vesicles; PRISMA-ScR, Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews. PRISMA-ScR flow diagram of the study-selection process. The diagram summarizes study identification through database and citation searching, removal of duplicate and ineligible records, title and abstract screening, full-text eligibility assessment, and final inclusion. A total of 99 studies were included through database searching and 15 through citation searching, resulting in 114 studies included in the review. Data Extraction and Classification by Primary Research Purpose A prespecified data extraction form was used to extract structured information from the included studies. The extracted data primarily comprised: 1) basic bibliographic information, including the first author, year of publication, and country or region of the first author’s affiliation; 2) study characteristics, including study type, disease context or model, EV source, and recipient cells or target tissues; 3) mechanistic information, including key cargo molecules, major functional pathways, and the corresponding bone-related effects; and 4) study findings and assessment information, including the main findings, bone-related outcomes, evidence-strength grade, and translational-stage classification. For participant-level human studies, additional data were extracted on sample type and particle source; the definition of OP or bone status; primary clinical endpoints; methods used to control for confounding factors; whether independent external validation was performed; whether receiver operating characteristic (ROC) analysis or other assessments of discriminatory performance were conducted; whether incremental clinical value beyond DXA, FRAX, or existing bone turnover markers was evaluated; and whether participant-derived EVs were used for functional validation. Given the substantial heterogeneity among the included studies in terms of disease context, EV source, isolation and characterization methods, key cargoes, target cells, and outcome measures, no quantitative synthesis was performed. Instead, the evidence was integrated using a combination of narrative synthesis and evidence mapping. Based on the primary research question and core conclusions of each study, the included evidence was classified as mechanism-focused, biomarker-focused, or intervention/delivery-focused. This classification was used to systematically present the distribution of evidence, predominant modes of action, completeness of mechanistic evidence chains, maturity of clinical or nanomedicine translation, and key evidence gaps within each category. Framework for Assessing Evidence Strength and Translational Stage To improve comparability across heterogeneous studies, each study was classified along two independent dimensions: evidence strength and translational stage. Evidence strength was used to assess the extent to which the evidence chain supporting the main study conclusion progressed from phenotypic observation and mechanistic support to attributional validation. Translational stage was used to assess the highest level of investigation that directly supported the main conclusion. Both assessments were based on the primary research question and core conclusion of each study. Evidence strength was classified into three levels: A, B, and C. Level A represented descriptive or associative evidence, primarily demonstrating differences in or associations between EVs or EV-like particles and bone-related phenotypes. Level B represented mechanism-supporting evidence, in which evidence regarding candidate cargoes, signaling pathways, recipient-cell interventions, or concordance between distribution and function was provided in addition to phenotypic effects, but remained insufficient to establish EV dependence, cargo attribution, or endogenous source attribution. Level C represented strengthened attributional evidence and was further designated according to the level of attribution supported: EV dependence (C-EV), cargo attribution (C-Cargo), or in vivo endogenous source attribution (C-Endo). The three Level C attribution dimensions could be assigned individually or in combination; however, they were not interchangeable, and evidence for one dimension was not considered sufficient to infer the presence of the others. Translational stage was classified from T0 to T3. T0 indicated that the main conclusion was supported only by in vitro or ex vivo experiments. T1 indicated that the main conclusion was supported by in vivo animal experiments. T2 indicated that the main conclusion was directly supported by participant-derived samples from individuals with clearly defined clinical status and phenotypes. T3 indicated that the research had progressed to clinical trials or clinical application, or that substantive evidence of industrial development was available. The use of human-derived cells, tissues, or EVs indicated only the origin of the study materials. When their functions were evaluated exclusively in vitro or in animal models, the studies were classified as T0 or T1, respectively, rather than T2. Each study was assigned the highest translational stage that directly supported its primary research question and core conclusion. The detailed operational assessment criteria are presented in Table 1. This framework was developed as an analytical tool for organizing and comparing evidence in the present scoping review and was not intended to replace conventional assessments of study quality or risk of bias. Results Overall Characteristics and Evidence Distribution of the Included Studies The database searches yielded 3952 records, including 1070 from PubMed, 1808 from Web of Science Core Collection, and 1074 from EMBASE. An additional 15 records were identified through citation searching. Before formal screening, 1346 records were removed, including 1325 duplicates, 5 standalone errata or correction notices, and 16 retraction notices. The remaining 2606 records underwent title and abstract screening, of which 1968 were excluded. A total of 638 reports underwent full-text retrieval and eligibility assessment, and the full texts of all reports were successfully obtained. After full-text assessment, 539 reports were excluded, primarily because the study object or target tissue did not meet the inclusion criteria (n = 224), a non-OP model outside the scope of this review was used (n = 237), or the EV source did not meet the inclusion criteria (n = 78). Ultimately, 99 studies were included through database searching and an additional 15 through citation searching, resulting in a total of 114 included studies. Based on the country or region of the first author’s affiliation, the largest number of studies originated from China, with 94 studies accounting for 82.5% of all included studies. This was followed by South Korea, with 8 studies (7.0%), and Japan, with 4 studies (3.5%). The remaining studies originated from Brazil (n = 3), Italy (n = 2), and the United States, Iran, and Germany (n = 1 each). Before consensus discussions, the two reviewers assigned the same evidence-strength grade to 109 studies, corresponding to an observed agreement of 95.6%; their assessments differed for the remaining 5 studies. The linearly weighted Cohen’s κ for evidence-strength grading was 0.93, indicating high inter-rater agreement. For translational stage, the two reviewers agreed on the classification of 111 studies, corresponding to an observed agreement of 97.4%, and disagreed on 3 studies. The linearly weighted Cohen’s κ for translational-stage classification was 0.95, likewise indicating high inter-rater agreement. Based on the primary research purpose of each study, the included studies were classified as mechanism-focused, biomarker-focused, or intervention/delivery-focused. Of the 114 studies, 39 were mechanism-focused, accounting for 34.2%; 22 were biomarker-focused, accounting for 19.3%; and 53 were intervention/delivery-focused, accounting for 46.5%. The included studies involved a broad range of EV and EV-like particle sources. Classification was based on the primary purpose of each study, whereas body fluids such as blood and urine primarily represented sampling media for EVs or routes through which donor cells were obtained. According to the evidence-strength grading framework, 55 studies were classified as Level C, accounting for 48.2%; 50 were classified as Level B, accounting for 43.9%; and 9 were classified as Level A, accounting for 7.9%. With respect to translational stage, T1 studies were the most common, with 76 studies accounting for 66.7%, followed by 24 T0 studies (21.1%) and 14 T2 studies (12.3%). No eligible published evidence at the T3 stage of clinical or industrial translation was identified in the bibliographic databases searched. After study selection, the included studies were systematically classified and summarized according to their primary research purpose, evidence strength, and translational stage, as shown in Figure 3.Figure 3Distribution of the included studies by primary research purpose, evidence strength, and translational stage, and an integrated evidence map. (A) Distribution by primary research purpose: mechanistic evidence (n = 39), biomarker evidence (n = 22), and intervention and delivery evidence (n = 53). (B) Distribution by evidence strength: Level A (n = 9), Level B (n = 50), and Level C (n = 55). (C) Distribution by translational stage: T0 (n = 24), T1 (n = 76), and T2 (n = 14); no eligible T3 evidence was identified in the bibliographic databases searched. (D) Integrated evidence map by evidence category and translational stage. Bubble size indicates the number of studies, bubble color indicates evidence strength, and y-axis label colors distinguish the three primary research purposes.Mixed charts showing study distribution by purpose, evidence strength, translational stage and a bubble map.Image A: Donut chart showing study type distribution. Slices: 46.5% Mechanistic evidence (n=39), 34.2% Biomarker evidence (n=22), 19.3% Intervention and delivery evidence (n=53). Image B: Bar chart of evidence strength. Categories A (9 studies, 7.9%), B (50 studies, 43.9%), C (55 studies, 48.2%). Image C: Bar chart of translational stage. Categories T0 (24 studies, 21.1%), T1 (76 studies, 66.7%), T2 (14 studies, 12.3%). No T3 evidence found. Image D: Bubble chart with translational stage (T0-T3) and evidence category. Categories: Immune cell, Muscle, Adipose, Vascular, Microbe, Participant level human mechanistic, Circulating blood, Urine stem cell, Native vesicles, Vesicles from perinatal tissues/pluripotent stem cells, Animal vesicles, Plant vesicles, Engineered vesicles. Dashed lines at T0-T3. Legend: Evidence strength (A, B, C), Number of studies (bubble sizes 1, 2, 5, 10, 14). Largest bubbles at T1 for Engineered vesicles and T2 for Circulating blood.Abbreviations: EVs, extracellular vesicles; T0, in vitro or ex vivo evidence; T1, in vivo animal evidence; T2, participant-level human evidence; T3, clinical or industrial translational evidence. Distribution of the included studies by primary research purpose, evidence strength, and translational stage, and an integrated evidence map. (A) Distribution by primary research purpose: mechanistic evidence (n = 39), biomarker evidence (n = 22), and intervention and delivery evidence (n = 53). (B) Distribution by evidence strength: Level A (n = 9), Level B (n = 50), and Level C (n = 55). (C) Distribution by translational stage: T0 (n = 24), T1 (n = 76), and T2 (n = 14); no eligible T3 evidence was identified in the bibliographic databases searched. (D) Integrated evidence map by evidence category and translational stage. Bubble size indicates the number of studies, bubble color indicates evidence strength, and y-axis label colors distinguish the three primary research purposes. Some studies used human-derived materials, but their main conclusions were supported only by in vitro or animal experiments; these studies were therefore classified as T0 or T1 according to the prespecified criteria. Among Level C studies, cargo-attribution evidence was relatively common, whereas rigorous evidence of in vivo endogenous source attribution was less frequent. Studies that conducted quantitative biodistribution assessments, independent external population validation, or product-level quality-control evaluations were also uncommon. As a tabular complement to the evidence map, Table 2 provides a concise summary of the available evidence according to primary research purpose. It summarizes the principal source or sample dimensions, core evidence, major targets or cargoes, current evidence base and translational maturity, and key gaps limiting further translation across mechanism-, biomarker-, and intervention/delivery-focused studies. Detailed characteristics of all included studies—including EV or EV-like particle source and donor state, recipient cells or clinical samples, experimental models, key cargoes or pathways, evidence grades, translational stages, and bibliographic information—are provided in Supplementary Tables S1–S14.Table 2Purpose-Stratified Evidence Maturity and Major Translational Gaps for Non-Bone-Derived EVs and EV-Like ParticlesSource/Sample AxisMajor EvidenceMain Targets/CargosCurrent MaturityKey Translational GapsMechanism-focused studies, n=39. Immune-, muscle-, adipose-, vascular-, and microbial-derived EVs, together with participant-derived or source-attributed human EV evidenceEVs generated under different donor states modulate osteogenesis, osteoclastogenesis, osteogenic–adipogenic fate, oxidative stress, inflammation, and cell-death programs. A subset of studies strengthened mechanistic attribution through EV-dependence tests, cargo perturbation, or source tracingPredominantly BMSCs, osteoblasts, osteocytes, and osteoclast-lineage cells. miRNAs are the most frequent cargoes, with additional proteins, circRNAs, metabolites, and inflammatory, redox, and metabolic pathwaysPredominantly preclinical. T0, n=15; T1, n=22; T2, n=2. Evidence level A, n=4; B, n=16; C, n=19Rigorous endogenous source-to-bone tracing remains uncommon; EV dependence, cargo attribution, and endogenous source attribution are rarely established together; longitudinal human mechanistic validation is lackingBiomarker-focused studies, n=22. Predominantly circulating plasma/serum EVs, together with preclinical circulating-EV studies and urine-related EV sourcesHuman studies associate EV abundance, surface features, and miRNA, tRF, protein, lncRNA/circRNA cargoes with BMD, T-scores, and bone-turnover markers. Selected candidates have undergone ROC assessment or functional testing using participant-derived EVsCandidate signals include EV-associated miRNAs, tRFs, proteins, and non-coding RNAs; major clinical endpoints include BMD, T-score, CTX, PINP, and OCN. A minority of studies further attribute circulating signals to hepatic, neuronal, or microbial sourcesGreatest representation of participant-level evidence. T0, n=1; T1, n=9; T2, n=12. Evidence level A, n=4; B, n=6; C, n=12; no T3 evidenceTissue origin of circulating EVs is usually unresolved; pre-analytical and isolation procedures remain heterogeneous; confounder adjustment, independent external validation, and longitudinal prediction are limited; incremental value beyond DXA, FRAX, or established bone-turnover markers remains unprovenIntervention/delivery-focused studies, n=53. Natural adult host-cell EVs, perinatal/pluripotent stem-cell EVs, animal- and plant-derived EV-like particles, and engineered or bone-targeted EV platformsAdministration of natural or engineered EVs/EV-like particles promotes osteogenesis, suppresses osteoclastogenesis, or improves bone phenotypes in multiple animal models. Engineered studies additionally evaluate therapeutic RNA/protein loading, surface functionalization, bone targeting, and biodistributionPredominantly BMSCs and osteoblast- and osteoclast-lineage cells. Functional payloads include miRNAs, proteins/growth factors, natural bioactive molecules, and engineered siRNAs; targeting strategies include bone-binding peptides and cell- or tissue-homing modificationsAlmost exclusively preclinical. T0, n=8; T1, n=45; no T2 or T3 studies. Evidence level A, n=1; B, n=28; C, n=24Human efficacy and safety evidence is absent; dose–response relationships, quantitative biodistribution, and potency assays remain insufficient; batch consistency, scalable manufacturing, GMP-compatible quality control, and long-term safety are unresolved; exogenous EV-like particles also require clearer active-component attribution and regulatory classificationNotes: Studies were grouped according to their primary research purpose as mechanism-focused, biomarker-focused, or intervention/delivery-focused. Evidence levels A–C indicate progressively strengthened support for the principal study conclusion, from phenomenological or associative evidence to mechanistic support and strengthened attribution. Translational stages T0–T3 indicate the highest level of evidence directly supporting the principal study conclusion, ranging from in vitro/ex vivo evidence to animal, participant-level, and clinical or industrial evidence. Blood and urine were treated primarily as sampling media unless a specific cellular or tissue origin was established.Abbreviations: BMD, bone mineral density; BMSCs, bone marrow mesenchymal stem cells; circRNA, circular RNA; CTX, C-terminal telopeptide of type I collagen; DXA, dual-energy X-ray absorptiometry; EVs, extracellular vesicles; FRAX, Fracture Risk Assessment Tool; GMP, Good Manufacturing Practice; lncRNA, long non-coding RNA; miRNA, microRNA; OCN, osteocalcin; PINP, procollagen type I N-terminal propeptide; ROC, receiver operating characteristic; siRNA, small interfering RNA; tRF, tRNA-derived fragment. Table 2Purpose-Stratified Evidence Maturity and Major Translational Gaps for Non-Bone-Derived EVs and EV-Like ParticlesSource/Sample AxisMajor EvidenceMain Targets/CargosCurrent MaturityKey Translational GapsMechanism-focused studies, n=39. Immune-, muscle-, adipose-, vascular-, and microbial-derived EVs, together with participant-derived or source-attributed human EV evidenceEVs generated under different donor states modulate osteogenesis, osteoclastogenesis, osteogenic–adipogenic fate, oxidative stress, inflammation, and cell-death programs. A subset of studies strengthened mechanistic attribution through EV-dependence tests, cargo perturbation, or source tracingPredominantly BMSCs, osteoblasts, osteocytes, and osteoclast-lineage cells. miRNAs are the most frequent cargoes, with additional proteins, circRNAs, metabolites, and inflammatory, redox, and metabolic pathwaysPredominantly preclinical. T0, n=15; T1, n=22; T2, n=2. Evidence level A, n=4; B, n=16; C, n=19Rigorous endogenous source-to-bone tracing remains uncommon; EV dependence, cargo attribution, and endogenous source attribution are rarely established together; longitudinal human mechanistic validation is lackingBiomarker-focused studies, n=22. Predominantly circulating plasma/serum EVs, together with preclinical circulating-EV studies and urine-related EV sourcesHuman studies associate EV abundance, surface features, and miRNA, tRF, protein, lncRNA/circRNA cargoes with BMD, T-scores, and bone-turnover markers. Selected candidates have undergone ROC assessment or functional testing using participant-derived EVsCandidate signals include EV-associated miRNAs, tRFs, proteins, and non-coding RNAs; major clinical endpoints include BMD, T-score, CTX, PINP, and OCN. A minority of studies further attribute circulating signals to hepatic, neuronal, or microbial sourcesGreatest representation of participant-level evidence. T0, n=1; T1, n=9; T2, n=12. Evidence level A, n=4; B, n=6; C, n=12; no T3 evidenceTissue origin of circulating EVs is usually unresolved; pre-analytical and isolation procedures remain heterogeneous; confounder adjustment, independent external validation, and longitudinal prediction are limited; incremental value beyond DXA, FRAX, or established bone-turnover markers remains unprovenIntervention/delivery-focused studies, n=53. Natural adult host-cell EVs, perinatal/pluripotent stem-cell EVs, animal- and plant-derived EV-like particles, and engineered or bone-targeted EV platformsAdministration of natural or engineered EVs/EV-like particles promotes osteogenesis, suppresses osteoclastogenesis, or improves bone phenotypes in multiple animal models. Engineered studies additionally evaluate therapeutic RNA/protein loading, surface functionalization, bone targeting, and biodistributionPredominantly BMSCs and osteoblast- and osteoclast-lineage cells. Functional payloads include miRNAs, proteins/growth factors, natural bioactive molecules, and engineered siRNAs; targeting strategies include bone-binding peptides and cell- or tissue-homing modificationsAlmost exclusively preclinical. T0, n=8; T1, n=45; no T2 or T3 studies. Evidence level A, n=1; B, n=28; C, n=24Human efficacy and safety evidence is absent; dose–response relationships, quantitative biodistribution, and potency assays remain insufficient; batch consistency, scalable manufacturing, GMP-compatible quality control, and long-term safety are unresolved; exogenous EV-like particles also require clearer active-component attribution and regulatory classificationNotes: Studies were grouped according to their primary research purpose as mechanism-focused, biomarker-focused, or intervention/delivery-focused. Evidence levels A–C indicate progressively strengthened support for the principal study conclusion, from phenomenological or associative evidence to mechanistic support and strengthened attribution. Translational stages T0–T3 indicate the highest level of evidence directly supporting the principal study conclusion, ranging from in vitro/ex vivo evidence to animal, participant-level, and clinical or industrial evidence. Blood and urine were treated primarily as sampling media unless a specific cellular or tissue origin was established.Abbreviations: BMD, bone mineral density; BMSCs, bone marrow mesenchymal stem cells; circRNA, circular RNA; CTX, C-terminal telopeptide of type I collagen; DXA, dual-energy X-ray absorptiometry; EVs, extracellular vesicles; FRAX, Fracture Risk Assessment Tool; GMP, Good Manufacturing Practice; lncRNA, long non-coding RNA; miRNA, microRNA; OCN, osteocalcin; PINP, procollagen type I N-terminal propeptide; ROC, receiver operating characteristic; siRNA, small interfering RNA; tRF, tRNA-derived fragment. Purpose-Stratified Evidence Maturity and Major Translational Gaps for Non-Bone-Derived EVs and EV-Like Particles Notes: Studies were grouped according to their primary research purpose as mechanism-focused, biomarker-focused, or intervention/delivery-focused. Evidence levels A–C indicate progressively strengthened support for the principal study conclusion, from phenomenological or associative evidence to mechanistic support and strengthened attribution. Translational stages T0–T3 indicate the highest level of evidence directly supporting the principal study conclusion, ranging from in vitro/ex vivo evidence to animal, participant-level, and clinical or industrial evidence. Blood and urine were treated primarily as sampling media unless a specific cellular or tissue origin was established. Abbreviations: BMD, bone mineral density; BMSCs, bone marrow mesenchymal stem cells; circRNA, circular RNA; CTX, C-terminal telopeptide of type I collagen; DXA, dual-energy X-ray absorptiometry; EVs, extracellular vesicles; FRAX, Fracture Risk Assessment Tool; GMP, Good Manufacturing Practice; lncRNA, long non-coding RNA; miRNA, microRNA; OCN, osteocalcin; PINP, procollagen type I N-terminal propeptide; ROC, receiver operating characteristic; siRNA, small interfering RNA; tRF, tRNA-derived fragment. Mechanistic Evidence for the Involvement of Non-Bone-Derived EVs in OP-Related Bone Remodeling Mechanistic evidence primarily involved EVs derived from immune cells, muscle, adipose tissue, the vasculature, and microorganisms, together with a small number of participant-level human mechanistic studies. These studies mainly evaluated the effects of EVs from donors under different conditions on BMSCs, OBs, osteocytes, OCs, or bone marrow immune cells and examined their effects on osteogenesis, osteoclastogenesis, the balance between osteogenic and adipogenic differentiation, and the bone marrow microenvironment. The direction of the observed effects varied according to donor state and disease context (Figure 4).Figure 4Donor-state-dependent effects of non-bone-derived extracellular vesicles on bone remodeling. EVs derived from immune cells, muscle, adipose tissue, vascular cells, and microorganisms regulate the bone marrow microenvironment by acting on BMSCs, OBs/osteocytes, and OCPs/OCs. EVs produced under homeostatic, reparative, exercise-associated, anti-inflammatory, or commensal/probiotic conditions generally promote bone formation and inhibit bone resorption, whereas EVs associated with inflammation, aging, metabolic dysfunction, mechanical unloading, or pathogenic microorganisms tend to impair osteogenesis and enhance osteoclastogenesis. Green and red arrows indicate generally favorable and unfavorable effects on bone remodeling, respectively. Upward (↑) and downward (↓) arrows indicate an increase or promotion and a decrease or inhibition, respectively, in the indicated cellular process or bone-remodeling outcome.Diagram: extracellular vesicles impact bone remodeling via immune, muscle, adipose, vascular, microbe sources.The diagram illustrates how extracellular vesicles from various sources influence bone remodeling. Vesicles from immune cells, muscles, adipose tissue, vascular cells and microorganisms affect the bone marrow environment. Immune cell vesicles, in both anti-inflammatory and inflammatory states, impact stromal cells, osteoblasts, osteocytes, osteoclast progenitors and osteoclasts. Muscle vesicles are shown in healthy and aging states, while adipose tissue vesicles appear in homeostasis and metabolic dysfunction states. Vascular cell vesicles are depicted in homeostatic and dysfunction states and microbe vesicles originate from probiotic and pathogenic bacteria. Green arrows indicate promotion of bone formation, while red arrows show inhibition. The diagram covers processes like proliferation, osteogenic differentiation, osteocyte function maintenance, osteoclast differentiation and bone resorption, with arrows indicating increases or decreases in these processes.Abbreviations: BMSCs, bone marrow mesenchymal stem cells; EVs, extracellular vesicles; MSCs, mesenchymal stem cells; OBs, osteoblasts; OCPs, osteoclast precursors; OCs, osteoclasts. Donor-state-dependent effects of non-bone-derived extracellular vesicles on bone remodeling. EVs derived from immune cells, muscle, adipose tissue, vascular cells, and microorganisms regulate the bone marrow microenvironment by acting on BMSCs, OBs/osteocytes, and OCPs/OCs. EVs produced under homeostatic, reparative, exercise-associated, anti-inflammatory, or commensal/probiotic conditions generally promote bone formation and inhibit bone resorption, whereas EVs associated with inflammation, aging, metabolic dysfunction, mechanical unloading, or pathogenic microorganisms tend to impair osteogenesis and enhance osteoclastogenesis. Green and red arrows indicate generally favorable and unfavorable effects on bone remodeling, respectively. Upward (↑) and downward (↓) arrows indicate an increase or promotion and a decrease or inhibition, respectively, in the indicated cellular process or bone-remodeling outcome. Immune-Cell-Derived EVs Seven studies of immune-cell-derived EVs were included, all of which focused on macrophage-derived EVs released under different polarization, stress, or infection states. Of these studies, three were classified as T0 and four as T1; two studies each provided Level A and Level B evidence, whereas three provided Level C evidence, all of which were classified as C-Cargo. Detailed information on EV source and donor state, functional cargoes, recipient cells, study models, evidence grades, and translational stages is provided in Supplementary Table S1. The effects of macrophage-derived EVs on bone remodeling were strongly dependent on donor state. EVs associated with M2 polarization generally promoted osteogenesis or inhibited osteoclastogenesis,47,64,65 whereas EVs generated under M1 polarization, simulated microgravity, or infection more commonly suppressed osteogenesis or promoted bone loss.66–68 EVs derived from immune cells under infectious conditions generally exerted detrimental effects on bone. Exosomes derived from Mycobacterium tuberculosis-infected macrophages inhibited OB function and bone formation; intervention targeting the key miRNA partially restored osteogenic indices and BMD.69 Muscle-Derived EVs Ten studies of muscle-derived EVs were included, with five classified as T0 and five as T1; five studies provided Level B evidence and five provided Level C evidence. All five Level C studies provided C-Cargo evidence, including one study that also provided C-EV evidence. Detailed information on EV source and donor state, functional cargoes, target cells, study models, evidence grades, and translational stages is provided in Supplementary Table S2. The bone-regulatory effects of muscle-derived EVs were likewise strongly dependent on donor state. Under basal, mechanically stimulated, or exercise-related conditions, muscle-derived EVs generally promoted osteogenesis, inhibited osteoclastogenesis, or improved bone phenotypes.45,46,70–74 In addition, EVs released by Zhuang-Gu-Fang-treated myoblasts promoted the osteogenic differentiation of BMSCs, with corresponding osteogenesis-related changes observed in a senescence-accelerated mouse prone 6 (SAMP6) model of senile OP.75 By contrast, inflammatory stimulation or impaired muscle function attenuated or even reversed these bone-protective effects.76 EVs derived from models with muscle-specific Slo1 deficiency inhibited osteogenesis both in vitro and in vivo and were associated with reduced bone mass and deterioration of bone microarchitecture; interventions targeting the relevant cargo or downstream signaling partially alleviated these inhibitory effects.77 Adipose-Derived EVs Six studies of adipose-derived EVs were included, of which two were classified as T0 and four as T1; four studies provided Level B evidence and two provided Level C evidence, with both Level C studies classified as C-Cargo. Detailed information on EV source and donor state, functional cargoes, target cells, study models, evidence grades, and translational stages is provided in Supplementary Table S3. Under conditions involving cellular injury, glucocorticoid exposure, or metabolic abnormalities, adipose-derived mesenchymal stem cell-derived extracellular vesicles (MSC-EVs) primarily protected osteogenic-lineage cells, promoted osteogenesis, and inhibited osteoclastogenesis.36,78–81 By contrast, exosomes derived from mature adipocytes mainly regulated the balance between osteogenic and adipogenic differentiation of BMSCs. Exosomes released by adipocytes treated with Wen-Shen-Tong-Luo-Zhi-Tong Decoction promoted osteogenic differentiation, suppressed adipogenic differentiation, improved bone microarchitecture, and reduced bone marrow fat accumulation in ovariectomized (OVX) mice.20 Vascular-Cell-Derived EVs Eight studies of vascular-cell-derived EVs were included, of which five were classified as T0 and three as T1; one study provided Level A evidence, two provided Level B evidence, and five provided Level C evidence, with all five Level C studies classified as C-Cargo. Detailed information on EV source and donor state, functional cargoes, target cells, study models, evidence grades, and translational stages is provided in Supplementary Table S4. The bone-regulatory effects of vascular-cell-derived EVs showed marked dependence on vascular cell subtype and microenvironmental context. These EVs regulated BMSC fate, OB function, osteoclastogenesis, and bone–vascular coupling. Most EVs derived from or associated with endothelial cells, vascular smooth muscle cells, or pericytes promoted osteogenesis, inhibited osteoclastogenesis, or attenuated bone injury.82–87 By contrast, EVs derived from microvascular endothelial cells under mechanical unloading exerted the opposite effect and suppressed osteogenesis.88 In addition, endothelial progenitor cell-derived extracellular vesicles (EPC-EVs) promoted BMSC proliferation and colony formation while simultaneously inhibiting osteogenic differentiation.89 Microbe-Derived EVs Six studies of microbe-derived EVs were included, all of which were classified as T1; three studies provided Level B evidence and three provided Level C evidence. Among the three Level C studies, one provided C-Cargo evidence and two provided C-EV evidence. Detailed information on EV source and microbial state, functional cargoes, target cells, study models, evidence grades, and translational stages is provided in Supplementary Table S5. EVs derived from commensal or probiotic bacteria generally promoted osteogenesis, inhibited osteoclastogenesis, and improved bone phenotypes in animal models. EVs derived from Roseburia intestinalis, Akkermansia muciniphila, and Lactobacillus salivarius promoted the osteogenic differentiation of BMSCs or inhibited OC formation and improved bone mass and microarchitecture in OVX animal models.90–92 Only some studies further validated the relevant functional cargoes and mechanisms of action, whereas the specific bioactive components remained unidentified in the other studies. OMVs derived from Proteus mirabilis also exerted anti-osteoclastogenic and bone-protective effects and alleviated OVX- and inflammation-associated bone loss.93 By contrast, EVs derived from pathogenic bacteria primarily suppressed osteogenesis, promoted osteoclastogenesis, and aggravated bone loss. OMVs derived from Helicobacter pylori impaired the osteogenic functions of BMSCs and OBs, enhanced OC differentiation, and aggravated bone loss in OVX mice.94 EVs derived from Klebsiella variicola enhanced inflammatory responses and OC formation, resulting in inflammation-associated bone loss.95 Participant-Level Human Mechanistic Evidence Among the mechanism-focused studies, only two met the definition of T2 participant-level human evidence used in this review. One study compared the bone-regulatory effects of T-cell-derived exosomes from postmenopausal women with and without OP,96 whereas the other evaluated the osteogenesis-inhibitory effects of serum- and bone marrow-derived EVs from patients with systemic mastocytosis.97 The former provided Level A evidence, whereas the latter provided C-Cargo evidence. Both studies directly used participant-derived EVs from individuals with clearly defined clinical identities and bone status in OB functional assays and related the functional findings to DXA-defined bone status or patient bone phenotypes. Six additional studies used human-derived materials;81,82,84,85,90,91 however, their principal EV-related conclusions remained supported by in vitro or animal experiments and were therefore classified as T0 or T1 according to the prespecified translational-stage criteria. Details of the study designs and clinical assessments are provided in Supplementary Table S6. Biomarker Evidence for Circulating Biofluid EVs Circulating Blood EVs Twenty studies involving circulating blood EVs were included, comprising eight preclinical functional studies and 12 participant-level studies. Of these, one was classified as T0, seven as T1, and 12 as T2; four studies provided Level A evidence, six provided Level B evidence, and 10 provided Level C evidence. All 10 Level C studies provided C-Cargo evidence. Two studies additionally fulfilled C-EV criteria, of which one also fulfilled the prespecified C-Endo criterion. Supplementary Table S7 presents the eight preclinical functional studies and summarizes EV source and state, target cells, experimental models, candidate cargoes or pathways, functional effects, evidence grades, and translational stages. Supplementary Table S8 presents the 12 participant-level human biomarker studies and summarizes EV samples or candidate markers, participant characteristics and definitions of bone status, clinical outcomes, control of confounding factors, validation and incremental clinical value, functional testing of participant-derived EVs, and evidence classifications. Animal-Related Evidence Animal studies showed that the bone-regulatory effects of circulating EVs varied according to physiological, disease, and intervention states. Circulating exosomes from mice with lipopolysaccharide (LPS)-induced bone loss and rats with chronic-stress-induced depression inhibited the osteogenic function of BMSCs or OBs.98,99 By contrast, serum exosomes from young rats improved the osteogenic capacity of BMSCs from aged rats with OP,100 whereas serum exosomes obtained after Erxian Decoction treatment attenuated senescence-related phenotypes in BMSCs.101 Another study also demonstrated pro-osteogenic effects of serum exosomes.102 Overall, the composition and function of circulating EVs appeared to differ substantially across systemic physiological and pathological states. Human-Derived Preclinical Evidence EVs derived from the plasma of young individuals and from human umbilical cord blood primarily exerted bone-protective effects in cellular or animal models. Plasma exosomes from young individuals promoted BMSC osteogenesis and improved bone mass in OVX animals,103,104 whereas human umbilical cord blood EVs improved bone mass and bone microarchitecture in mice with senile OP.105 Because the principal EV-related conclusions of these studies remained supported by in vitro or animal experiments, they were classified as human-derived preclinical evidence rather than T2 participant-level human evidence. Human Evidence Participant-level studies first suggested that circulating EV cargoes may be associated with bone-mass stratification and the diagnosis of OP. Plasma exosomal miRNA profiles differed among postmenopausal women with different BMD categories, and several miRNAs and tRNA-derived fragments were identified as candidate discriminatory markers for postmenopausal osteoporosis (PMOP) or OP.48,51,106 These studies primarily used cross-sectional or case-control designs and focused on concurrent BMD or OP diagnosis; the longitudinal value of these markers for predicting subsequent bone loss or fracture risk was not evaluated. Some studies further demonstrated direct bone-cell-regulatory effects of participant-derived circulating EVs. The composition of circulating EVs was altered in individuals with low bone mass or OP, and these EVs promoted OC formation or inhibited osteogenesis and mineralization in ex vivo or cellular experiments.24,107,108 The functions of selected noncoding RNA cargoes were further supported by intervention experiments or animal models.109 Candidate EV cargoes associated with BMD or bone-turnover markers were also detected in individuals with Alzheimer’s disease, aging-related bone loss, OP, or depression. Some studies combined these human observations with experiments involving EVs from the corresponding cellular or organ sources to assess effects on osteogenic–adipogenic differentiation or OC formation.23,50,110–112 Overall, the available human evidence primarily supported cross-sectional associations between circulating EV cargoes and bone status and provided partial functional validation. However, source attribution, independent external validation, and longitudinal predictive value remained limited. EVs Derived from Urine-Derived Stem Cells Two studies of EVs derived from urine-derived stem cells were included, both of which were classified as T1 and provided C-Cargo evidence. In both studies, urine-derived stem cells were first isolated from urine and expanded in vitro, after which their secreted EVs were collected. Thus, urine primarily represented the route through which the donor cells were obtained rather than a medium from which EVs were directly sampled. Detailed information on EV source and state, target cells, study models, functional cargoes, evidence grades, and translational stages is provided in Supplementary Table S9. Human urine-derived stem cell-derived extracellular vesicles (USC-EVs) generally promoted osteogenesis, inhibited osteoclastogenesis, and improved bone mechanical properties. USC-EVs generally promote osteogenesis, inhibit osteoclastogenesis, and improve bone mechanical properties. Cell-based experiments supported the involvement of collagen triple helix repeat containing 1 (CTHRC1) and osteoprotegerin (OPG) transported by USC-EVs in their pro-osteogenic and anti-osteoclastogenic effects, respectively.113 In a model of diabetic osteoporosis (DOP), USC-EVs improved bone mineralization and bone-formation indices and reduced osteoclast activity, while cargo-intervention experiments supported the involvement of miR-26a-5p in these bone-protective effects.114 Intervention and Delivery Evidence for EVs and EV-Like Particles The major intervention sources, engineering strategies, routes of administration, and in vivo delivery processes are summarized in Figure 5.Figure 5Intervention and delivery framework for non-bone-derived extracellular vesicles and extracellular vesicle-like particles. (A) Major intervention sources, including adult cell-derived, perinatal tissue- and pluripotent stem cell-derived, animal-derived, plant-derived, and engineered EVs or EV-like particles. (B) Representative engineering and intervention strategies, including donor-cell preconditioning or genetic modification, therapeutic cargo loading, bone-targeting functionalization, enteric coating, and imaging-probe integration. (C) Routes of administration and in vivo delivery processes, including intravenous, local, and oral administration, followed by MPS-mediated clearance, tissue distribution, bone-tissue accumulation, bone-marrow retention, and target-cell uptake.Flowchart showing EV sources, modification steps and how they reach bone cells after dosing.Three-panel schematic labeled A to C. Panel A lists five origins of extracellular vesicles/EV-like particles with small icons: adult cell to derived, perinatal tissue and pluripotent stem cell to derived, animal-derived (eg, milk/seafood/insect), plant-derived (leaf/root/carrot) and engineered vesicles. Arrows lead to Panel B, which groups intervention methods into three boxes: (1) donor-cell preparation (preconditioning, genetic modification, transfection), (2) loading/enriching therapeutic contents (miRNA/siRNA, peptides, other bioactive cargo) and (3) formulation tweaks (bone-targeting surface features, enteric coating and imaging labels). Panel C depicts administration options intravenous infusion, local injection near bone and oral delivery through the gut followed by a pathway of in-body fates: clearance by phagocyte systems, distribution to organs, accumulation in bone, retention in marrow and uptake by bone-related cells (BMSCs, osteoblasts, osteocytes and osteoclasts).Abbreviations: BMSCs, bone marrow mesenchymal stem cells; CLEC11A, C-type lectin domain family 11 member A; EVs, extracellular vesicles; miRNA, microRNA; MPS, mononuclear phagocyte system; OBs, osteoblasts; OCs, osteoclasts; PSCs, pluripotent stem cells; siRNA, small interfering RNA. Intervention and delivery framework for non-bone-derived extracellular vesicles and extracellular vesicle-like particles. (A) Major intervention sources, including adult cell-derived, perinatal tissue- and pluripotent stem cell-derived, animal-derived, plant-derived, and engineered EVs or EV-like particles. (B) Representative engineering and intervention strategies, including donor-cell preconditioning or genetic modification, therapeutic cargo loading, bone-targeting functionalization, enteric coating, and imaging-probe integration. (C) Routes of administration and in vivo delivery processes, including intravenous, local, and oral administration, followed by MPS-mediated clearance, tissue distribution, bone-tissue accumulation, bone-marrow retention, and target-cell uptake. Native EVs Derived from Adult Cells Twelve studies were included in this section, of which one was classified as T0 and 11 as T1, with no T2 or T3 evidence. Evidence levels comprised one Level A, four Level B, and seven Level C studies, all classified as C-Cargo. Three human-derived preclinical studies used EVs from human skeletal muscle cells, microvascular endothelial cells, or adipose-derived stem cells; because their conclusions were validated in animal models, these studies remained classified as T1. Supplementary Table S10 summarizes EV source and donor state, target cells, study models, functional cargoes or pathways, administration and combination strategies, evidence grades, translational stages, and bone-related effects. Intervention studies of native EVs derived from adult cells primarily involved particles of muscle,115–118 immune,119,120 adipose,49,121 and vascular origin.122–124 In most studies, native EVs were isolated directly or collected after donor-cell preconditioning and subsequently administered exogenously. Improvements in bone mass, bone microarchitecture, or bone-remodeling indices were observed in animal models of OP. Some studies also sought to achieve dual therapeutic effects by simultaneously correcting the primary extra-skeletal pathological environment and abnormal bone remodeling. For example, combined administration of intestinal organoid-derived EVs and intestinal organoids restored the intestinal mucosal barrier, reduced inflammation, and concurrently ameliorated inflammatory-bowel-disease-associated bone loss.125 EVs Derived from Perinatal Tissues and Pluripotent Stem Cells Eight studies of EVs derived from perinatal tissues and pluripotent stem cells were included, of which one was classified as T0 and seven as T1; six studies provided Level B evidence and two provided Level C evidence, with both Level C studies classified as C-Cargo. All eight studies constituted human-derived preclinical evidence. Seven included in vivo validation in mouse or rat models of OP or age-related bone loss and were therefore classified as T1, whereas one involved only in vitro functional experiments using human preosteoblasts and was classified as T0. Detailed information on EV source and treatment state, target cells, study models, functional cargoes or pathways, evidence grades, and translational stages is provided in Supplementary Table S11. MSC-EVs derived from the umbilical cord and Wharton’s jelly constituted the main body of evidence for perinatal-tissue-derived EVs. Wharton’s jelly mesenchymal stem cell-derived extracellular vesicles (WJ-MSC-EVs) reached the distal femur, enhanced local osteogenic activity, and improved bone mass and trabecular architecture in OVX mice.126 Human umbilical cord mesenchymal stem cell-derived exosomes (hUCMSC-Exos) obtained under conventional culture or osteogenic-induction conditions both alleviated bone loss in OVX mice; the former more strongly promoted cell proliferation, whereas the latter exerted greater effects on osteogenic differentiation and mineralization.127 In models of glucocorticoid-induced osteoporosis (GIOP), hUCMSC-Exos also attenuated injury to human BMSCs, restored their osteogenic capacity, and improved bone microarchitecture in mice.128 Some studies provided more direct evidence for functional cargo attribution. hUCMSC-EVs promoted bone formation, reduced bone marrow adipogenesis, and inhibited bone resorption, thereby improving bone mass and strength in OVX, aging-related, and disuse OP models. Donor-cell cargo-intervention experiments supported the contribution of CLEC11A to their pro-osteogenic, anti-adipogenic, and anti-osteoclastogenic effects.129 Another study of disuse OP showed that hUCMSC-Exos reduced BMSC apoptosis and improved trabecular architecture, with cargo-intervention experiments supporting the involvement of miR-1263 in these bone-protective effects.130 EVs derived from amniotic membrane and amniotic fluid stem cells primarily promoted osteogenesis and protected osteogenic-lineage cells. EVs derived from hypoxia-preconditioned human amniotic membrane stem cells improved BMD, bone microarchitecture, and bone mechanical strength in OVX rats, with some effects showing dose dependence.131 Human amniotic fluid stem cell-derived EVs attenuated dexamethasone (Dex)-induced apoptosis and oxidative injury in preosteoblasts in vitro and restored osteogenic differentiation and mineralization; however, animal-level bone-phenotype validation was lacking.132 Pluripotent-stem-cell-derived EVs primarily restored the function of aged bone marrow stem cells. Long-term administration of human embryonic stem cell-derived small extracellular vesicles (hESC-sEVs) improved bone mass and bone microarchitecture in aged mice and restored the proliferative and osteogenic capacities of senescent BMSCs.133 Animal-Derived EVs and EV-Like Particles Six studies of animal-derived EVs and EV-like particles were included, all of which were classified as T1 and provided Level B evidence. The study of medicinal-insect-derived EV-like particles also included in vitro functional validation using human bone marrow mesenchymal stem cells (hBMSCs) and therefore constituted preclinical evidence incorporating human-cell experiments. Detailed information on particle source and state, target cells, study models, bioactive components or related pathways, evidence grades, and translational stages is provided in Supplementary Table S12. Milk-derived EVs primarily inhibited osteoclastogenesis, maintained osteocyte homeostasis, and attenuated metabolic bone injury. Bovine milk-derived EVs inhibited OC differentiation and improved femoral bone mass and microarchitecture in OVX mice.134 Orally administered bovine milk-derived EVs also increased the numbers of OBs and osteocytes, reduced OC formation, and improved bone microarchitecture and femoral mechanical properties in OVX models and models of bone loss associated with a highly refined carbohydrate diet.135,136 EVs derived from the mantles of marine shellfish primarily promoted osteogenesis and inhibited osteoclastogenesis and were able to reach bone tissue after oral administration. Oyster mantle-derived exosomes promoted OB differentiation, inhibited OC formation, and increased new bone formation and improved bone mass and trabecular architecture in femoral-defect and OVX animal models.137 Scallop mantle-derived exosomes likewise promoted OB proliferation, differentiation, and mineralization. After oral administration, these particles were distributed to bone tissue and improved BMD, bone microarchitecture, and bone-turnover indices in OVX mice.138 EV-like particles derived from the medicinal insect Steleophaga plancyi were internalized by hBMSCs and promoted their proliferation, osteogenic differentiation, and mineralization. After oral administration in an enteric-coated formulation, these particles increased bone mass and improved bone microarchitecture in OVX rats. Although the available findings implicated melatonin-associated activation of autophagy, the causal contributions of specific bioactive components and the underlying mechanisms require further validation.139 Plant-Derived EV-Like Particles Eleven studies of plant-derived EV-like particles were included, of which five were classified as T0 and six as T1; all provided Level B evidence. Four studies included preclinical validation using human cells and involved particles derived from Pueraria lobata, Rhizoma Drynariae, gold kiwifruit, or Epimedium brevicornu. Except for the gold-kiwifruit study, which included only in vitro experiments, the other three studies were further validated in OVX animal models. Detailed information on particle source and state, target cells, study models, candidate bioactive components or related pathways, evidence grades, and translational stages is provided in Supplementary Table S13. Plant-derived EV-like particles generally promoted osteogenic differentiation and mineralization. Particles derived from yam, Panax notoginseng, the brown alga Sargassum fulvellum, or apple enhanced the proliferation, osteogenic differentiation, or mineralization of BMSCs or OB-lineage cells. Yam-derived particles additionally improved BMD and bone-formation-related phenotypes in OVX mice, whereas the principal functional evidence for particles derived from P. notoginseng, S. fulvellum, and apple remained based on in vitro experiments.140–143 Some plant-derived particles also regulated both bone formation and bone resorption. Plum-derived particles promoted osteogenesis while suppressing OC formation and activation in vitro, indicating potential bidirectional regulation of bone remodeling,144 whereas P. lobata-derived particles promoted hBMSC osteogenesis and improved bone mass and bone microarchitecture in OVX rats.145 The bone-protective effects of other plant-derived particles were accompanied by attenuation of cellular injury or improvement of adverse microenvironmental conditions, suggesting that these processes may contribute to their actions. Gold-kiwifruit-derived nanovesicles reduced oxidative stress in injured human BM-MSCs, restored their proliferative, migratory, and osteogenic capacities, and inhibited adipogenic differentiation.146 Carrot-derived nanovesicles aggregated within the acidic microenvironment associated with OP, attenuated oxidative stress and mitochondrial injury in OBs, and reduced bone loss in osteoporotic mice.147 Particles showing preferential accumulation in bone tissue further demonstrated potential for in vivo intervention. Rhizoma Drynariae-derived nanovesicles accumulated in bone, promoted hBMSC proliferation and osteogenesis, and improved bone mass and microarchitecture in OVX mice; these effects may have been partly related to their naringin content.148 Morinda officinalis-derived EV-like particles likewise showed femoral accumulation and bone-protective activity, although their in vitro effects primarily involved promotion of cell proliferation, with relatively limited direct effects on osteogenic differentiation.149 Epimedium brevicornu-derived EV-like particles accumulated in osteoporotic femora, improved the intraosseous vascular environment, and attenuated bone loss in OVX mice, suggesting that modulation of the vascular microenvironment may contribute to their bone-protective effects.150 Engineered EVs and EV-Like Particles Sixteen studies of engineered EVs and EV-like particles were included, of which one was classified as T0 and 15 as T1; one study provided Level B evidence and 15 provided Level C evidence, including 14 classified as C-Cargo and one as C-EV. Engineering strategies primarily included donor-cell preconditioning or genetic modification, loading of therapeutic nucleic acids and proteins or peptides, membrane-surface functionalization, and targeting of bone tissue or specific bone-cell populations. Six studies constituted human-derived preclinical evidence and involved EVs derived from human adipose-derived MSCs, human induced pluripotent stem cell (iPSC)-derived MSCs, human umbilical vein endothelial cells (HUVECs), or HEK293/HEK293T cells. Because their principal conclusions were validated in animal models, these studies were classified as T1. Detailed information on EV or EV-like particle source, engineering strategy, target cells, functional cargoes, study models, evidence grades, and translational stages is provided in Supplementary Table S14. Engineered EVs from nonhuman host cells primarily used genetic modification of donor cells to increase the abundance of therapeutic cargoes. Exosomes derived from glucagon-like peptide 2 (GLP-2)-overexpressing macrophages inhibited OC differentiation and attenuated bone loss in OVX rats.151 Genetically modified adipose-derived stem cell EVs enhanced OB protection, inhibited inflammatory bone resorption, and improved bone phenotypes in models of secondary bone loss, including ankylosing-spondylitis-associated spinal OP and diabetic OP.152–154 Some studies further used membrane-surface modification and multicomponent loading to improve bone distribution, therapeutic cargo delivery, and imaging capacity. Engineered milk-derived EVs combined bone accumulation, regulation of bone remodeling, and MRI tracking, thereby demonstrating theranostic potential.155 Bone-targeting-peptide-modified Schwann-cell-derived EVs also showed increased bone accumulation and inhibition of OC formation, integrating native anti-osteoclastogenic activity with engineered bone-homing capacity.156 Engineering of microbe-derived vesicles primarily focused on bone targeting, recognition of specific bone-cell populations, and delivery of therapeutic cargoes. Bone-targeted bacterial OMVs or bacterial EVs showed enhanced accumulation in the bone microenvironment, promoted BMSC osteogenesis, or recognized osteoclast precursors (OCPs) and delivered anti-osteoclastogenic peptides or nucleic acids.157–159 Other platforms simultaneously displayed bone-homing molecules and osteogenic proteins or introduced bone-targeting modifications while retaining native functional cargoes, thereby integrating targeted delivery with bone-regulatory activity.160,161 Human-derived engineered EVs primarily focused on bone targeting, therapeutic cargo enrichment, and interorgan delivery. Bone-targeted human iPSC-MSC-derived exosomes loaded with siRNA preferentially acted on OB-lineage cells, promoted osteogenesis, suppressed bone resorption, and improved angiogenic–osteogenic coupling.44 HUVEC-derived exosomes were mainly engineered to increase functional protein or miRNA content and thereby enhance the bone-protective effects of native endothelial EVs. These studies demonstrated enhanced angiogenic–osteogenic coupling, inhibition of OC formation, or simultaneous promotion of osteogenesis and suppression of osteoclastogenesis.162–164 In addition, engineered exosomes derived from HEK293/HEK293T cells were used to deliver siRNA to glomerular mesangial cells and were evaluated for their effects on impaired osteogenesis and implant osseointegration associated with the kidney–bone axis in diabetic kidney disease.165 From Multi-Axis Mechanistic Evidence to EV-Based Nanomedicine Translation: Evidence Chains, Product Development, and Clinical Validation Taken together, the available evidence on non-bone-derived EVs and EV-like particles remained predominantly preclinical, with marked differences in evidential maturity across mechanistic, biomarker, and intervention and delivery research. On the basis of this evidence structure, Figure 6 summarizes the principal pathways and key bottlenecks extending from mechanistic validation to clinical translation.Figure 6Translational maturity framework and purpose-oriented research pathways for non-bone-derived extracellular vesicles and extracellular vesicle-like particles. (A) Progressive requirements from T0 in vitro/ex vivo evidence through T1 animal in vivo evidence and T2 participant-level evidence to T3 clinical and industrial translation. (B) Purpose-specific development pathways for mechanistic, biomarker, and intervention/delivery research, highlighting source and cargo attribution, longitudinal and external validation, biodistribution and dose optimization, potency and manufacturing development, and clinical efficacy, safety, and utility.A diagram of translational maturity for non-bone-derived EVs, showing stages from T0 to T3 and research pathways.A diagram shows the growing translational maturity of non-bone-derived extracellular vesicles (EVs) and EV-like particles. Stages include: T0 - in vitro/ex vivo evidence, focusing on EV characterization, cellular uptake, cargo screening and EV-dependence. T1 - animal in vivo evidence, addressing dose, biodistribution, bone exposure, phenotype, microarchitecture and safety. T2 - participant-level evidence, involving clinical cohorts, EV origin, stratification, validation and predictive value. T3 - clinical and industrial translation, covering GMP manufacturing, batch consistency, safety, pharmacodynamics, BMD, bone microarchitecture, fractures and pharmacovigilance. Research pathways include mechanistic, biomarker and intervention/delivery research, each with specific focus areas and validation processes.Abbreviations: BMD, bone mineral density; DXA, dual-energy X-ray absorptiometry; EVs, extracellular vesicles; GMP, good manufacturing practice; T0–T3, translational stages 0–3. Translational maturity framework and purpose-oriented research pathways for non-bone-derived extracellular vesicles and extracellular vesicle-like particles. (A) Progressive requirements from T0 in vitro/ex vivo evidence through T1 animal in vivo evidence and T2 participant-level evidence to T3 clinical and industrial translation. (B) Purpose-specific development pathways for mechanistic, biomarker, and intervention/delivery research, highlighting source and cargo attribution, longitudinal and external validation, biodistribution and dose optimization, potency and manufacturing development, and clinical efficacy, safety, and utility. Intersystem Communication in Multi-Axis Regulation of OP by Non-Bone-Derived EVs The effects of non-bone-derived EVs on OP-related bone remodeling are not governed by a single organ axis but involve the immune system, muscle, adipose tissue, the vasculature, and other extra-skeletal systems.27,166,167 EVs from different sources may act on common effector cells involved in bone remodeling, but the direction of their effects is influenced by donor state and disease context. The available evidence therefore supports interpreting non-bone-derived EVs as candidate mediators that transmit multisystem states into the bone-remodeling process, rather than as a single pathogenic or protective mechanism with a fixed direction of action.166,168,169 At the same time, substantial differences exist among studies in source attribution, cargo causality, and translational maturity.26,40,53,56,170,171 Mechanistic evidence therefore cannot be considered equivalent to evidence of therapeutic delivery. Before EVs are defined as candidate therapeutics, the field must first determine which evidence chains have been adequately established, which critical gaps remain, and which translational pathways are sufficiently mature to proceed to clinical validation.32 Evidence-Chain Gaps Determining Translational Maturity Gaps in Source Tracing Evidence of delivery capacity should be clearly distinguished from evidence of an endogenous mechanism. Biological activity observed after exogenous EV administration does not demonstrate that the same EV population naturally drives the development or progression of OP under physiological or pathological conditions.172 Mechanistic studies must further determine whether specific EVs are genuinely produced in vivo, whether they reach bone tissue, whether they achieve an effective level of exposure, and whether they exert functionally meaningful skeletal effects.50 Because EVs constitute highly heterogeneous particle populations, a change in a single cargo is insufficient to define their tissue of origin or overall function.173 Future studies should use cell-specific EV labeling, genetic tracing, inhibition of EV release, and conditional manipulation of donor cells to determine whether EVs derived from a specific organ or cell type genuinely participate in OP progression.34 Only after such evidence has been established can non-bone-derived EVs advance from associative signals to pathologically meaningful regulatory nodes or criteria for patient stratification.172 Importantly, opposing findings in the available literature are more appropriately interpreted as evidence of context-dependent EV effects than as evidence that a particular source axis is intrinsically protective or pathogenic. Within the immune axis, M2-macrophage-derived EVs generally promoted osteogenesis or inhibited osteoclastogenesis, whereas macrophage EVs generated under M1 polarization, simulated microgravity, or infection produced opposite effects.64,65,67–69 Within the muscle axis, EVs produced under basal, mechanically stimulated, or exercise-related conditions generally supported bone formation or inhibited bone resorption, whereas TNF-α stimulation or impaired muscle function attenuated or even reversed these effects.46,72–77 Within the vascular axis, most endothelial-cell- or pericyte-associated EVs promoted osteogenesis or inhibited osteoclastogenesis, whereas microvascular-endothelial-cell-derived EVs generated under mechanical unloading suppressed osteogenesis.82–88 These differences may be influenced simultaneously by donor-cell state, disease context, EV isolation and characterization methods, dose and route of administration, recipient-cell type, and outcome selection. Because most opposing findings were obtained from different experimental systems rather than from direct comparisons under standardized conditions, they should not yet be interpreted as demonstrating a confirmed biological “effect reversal”. A more cautious conclusion is that EVs represent a composite output of donor state and experimental context. Future studies should directly compare EVs derived from the same source under different states using standardized particle characterization, dosing, and disease models, together with validation of EV dependence, cargo intervention, and in vivo source tracing, to distinguish genuine state-dependent effects from apparent inconsistencies caused by methodological heterogeneity. Gaps in Causal Cargo Attribution Existing studies frequently use miRNAs or other differentially abundant cargoes to explain the effects of EVs on bone remodeling. However, EVs are composite signaling units comprising nucleic acids, proteins, lipids, metabolites, and membrane-associated molecules.31,169,172,174 Differential abundance of an individual cargo therefore does not directly demonstrate that it causally drives changes in osteogenesis, osteoclastogenesis, or osteogenic–adipogenic cell fate. Candidate cargoes identified solely through omics screening, differential-expression analysis, target prediction, or pathway correlations are more appropriately described as potential mechanistic leads than as established causal mediators.172 Stronger cargo attribution requires a more complete functional evidence chain. This includes demonstrating that the candidate cargo enters recipient cells within EVs, showing that knockdown, overexpression, or another intervention alters the EV-associated bone phenotype, and further using cargo reconstitution or pathway-rescue experiments to establish necessity or functional relevance.171,172 Only a subset of the studies included in this review performed relatively complete intervention and rescue experiments.45,64,67,70,97,115 Consequently, the strength of causal cargo attribution continued to vary markedly among studies. Bone Homing, Biodistribution, and Targeted Delivery: Key Barriers to Nanomedicine Translation Bone-homing capacity remains one of the major translational barriers facing non-bone-derived EVs. BMSC-derived EVs may retain some of the homing characteristics of their parent cells, whereas many non-bone-derived EVs lack validated determinants of bone homing, and their exposure within the bone marrow microenvironment has not been adequately quantified.175,176 Following systemic administration, a substantial proportion of EVs are cleared by the mononuclear phagocyte system, and only a small fraction reaches the bone marrow.44,177–179 This creates uncertainty regarding the effective dose, dosing frequency, duration of pharmacological activity, and safety.180 Bone-targeting ligands, membrane-surface functionalization, and other engineering strategies have been used to enhance EV accumulation in bone tissue.43,157–159,181 However, most studies still lack systematic pharmacokinetic analyses, quantitative tissue-distribution assessments, and measurements of intraosseous retention.43 Research on bone-related nanomaterials has also shown that combining components such as hydroxyapatite and chitosan can regulate osteoconductivity, interactions at the bone interface, and the local presentation of bioactive molecules.182–185 Dedicated experiments are needed to determine whether such composite strategies preserve EV integrity and biological activity while improving local exposure, sustained release, and therapeutic efficacy in bone tissue. Gaps Related to Population Heterogeneity Clinical translation must adequately account for the heterogeneity of the OP population. Existing studies have primarily used individual models based on OVX, aging, glucocorticoid exposure, diabetes, or disuse, each of which reflects only part of the complexity of real-world populations with OP.28,186 EV source composition, release levels, and cargo profiles may also be affected by age, menopausal status, inflammation, metabolic status, physical activity, renal function, and medication use.24,30,39 The absence of etiological and biological stratification may therefore reduce the reproducibility of EV-associated signals in clinical populations. Gaps in Product Heterogeneity and Standardization EV composition and function may be influenced by donor source, cell-culture conditions, preconditioning strategies, isolation and purification methods, and storage conditions, resulting in substantial batch-to-batch variation.39,179,187–189 EVs derived from plasma or culture systems may also be affected by co-isolated lipoproteins, protein aggregates, free nucleic acids, and culture-medium components.39,173 Standardization of therapeutic EVs therefore cannot rely solely on particle counts or routine characterization. It must also address product identity, purity, dose, mechanism-related potency, safety, and batch consistency.190,191 Product development for OP will additionally require bone-remodeling-related potency assays and scalable, batch-stable manufacturing and release procedures that comply with good manufacturing practice (GMP) requirements.186,191–194 The central objective of product standardization should shift from merely demonstrating the presence of EV particles to demonstrating that different product batches maintain reproducible compositional characteristics, functional potency, and safety attributes.190,195 Priority Translational Pathways for Non-Bone-Derived EVs in OP Biomarker Pathway Among the potential translational directions, EV biomarkers may enter application-oriented validation earlier than therapeutic EVs. Biomarker studies can use readily accessible samples such as plasma, serum, and urine without having to address the therapeutic-development challenges of bone targeting, in vivo exposure, or the long-term safety of repeated administration.24,196 The most clinically relevant evidence currently comes from biofluid EVs, particularly altered miRNAs, tRNA-derived fragments, and protein cargoes in plasma or serum EVs from patients with OP or osteopenia and from healthy controls.51,107 These data do not yet establish independent diagnostic value for any single EV cargo, but they suggest that circulating EV cargoes may reflect low bone mass, abnormal bone remodeling, and fracture-related states.197 Future studies should validate EV-based measures in multicenter longitudinal cohorts and integrate them with BMD, bone-turnover markers, high-resolution peripheral quantitative computed tomography (HR-pQCT)-derived bone microarchitecture, fragility-fracture history, and treatment response.30,198,199 EV biomarkers must demonstrate incremental value beyond established clinical tools such as DXA, FRAX, procollagen type I N-terminal propeptide (PINP), and C-terminal telopeptide of type I collagen (CTX) rather than merely showing statistically significant differences between groups.48,198,199 Under reproducible procedures for sample processing, EV isolation, and particle characterization, candidate biomarkers should undergo ROC analysis, calibration assessment, external validation, reclassification analysis, and decision-curve analysis.39,48 Because plasma and serum EVs are mixed particle populations, studies should also report sample type, anticoagulant, centrifugation conditions, isolation method, control of lipoprotein contamination, and source-associated markers in detail.39,48 Given the marked heterogeneity of OP, EV biomarkers may be most appropriately incorporated into composite models integrating particle characteristics, source information, multi-omics cargo profiles, and clinical phenotypes for patient stratification, risk prediction, and treatment monitoring.30,200,201 Mechanistic Phenotyping and Participant Stratification The heterogeneity of OP means that BMD, age, and fracture history alone cannot fully explain differences among patients in bone turnover, bone microarchitecture, or treatment response.25,30,202 The translational value of non-bone-derived EVs may therefore extend beyond drug development to the identification of the principal pathological drivers operating in different patient groups.56,186,203 For example, muscle-derived EVs may be particularly relevant to sarcopenia- or disuse-associated bone loss,70,117 immune-cell-derived EVs to inflammation- or immune-activation-associated bone loss,67,119,120 adipose- and metabolism-associated EVs to OP associated with obesity, diabetes, or increased bone marrow adiposity,20,36,154 and vascular-cell-derived EVs to patients with vascular injury or impaired bone–vascular coupling.83,123,162 Compared with an individual differentially abundant cargo, integration of EV source, cargo characteristics, target-cell effects, and clinical phenotypes may be more suitable for patient stratification and selection of mechanistic endpoints.39,173,204 Therapeutic EV Product Types for Priority Development On the basis of the available evidence, adipose-derived MSC-EVs and perinatal-tissue-derived MSC-EVs have relatively well-defined sources, offer some feasibility for donor-cell expansion, and have been evaluated in a comparatively large number of preclinical bone-protection studies. They may therefore represent priority candidates for the next stage of product development.53,192,193 Muscle- and vascular-cell-derived EVs are currently more suitable as biological leads for disease-mechanism research and engineering design.71,117,164 At the present stage, simultaneous development of a large number of EV products from different sources is unlikely to be appropriate. Instead, translational validation should prioritize a limited number of candidate systems that already have clearly defined sources, relatively stable bone-related potency, and an acceptable preliminary safety basis.191,203 Translational Pathways for Intervention and Delivery Using Exogenous Vesicle-Like Particles Microbe-, plant-, and animal-derived vesicle-like particles differ from mammalian-cell-derived EVs in biogenesis, composition, and regulatory characteristics and therefore require distinct product positioning.22 Microbe-derived particles may be developed as microbiota-associated functional carriers or engineered delivery platforms, but inflammatory activity and biosafety risks must be prioritized.91,92,158,159,195,205,206 Plant-derived particles more closely resemble natural intervention resources or oral-delivery platforms and require further clarification of their bioactive components, in vivo exposure, and quality standards.39,140,145,148,149,207 Animal-derived EVs have the advantage of food-based accessibility, but their oral stability, long-term safety, and effects on both bone mass and bone quality require evaluation.134,137–139,208,209 At present, these exogenous particles are more appropriately regarded as a preclinical pipeline of natural nanointerventions and delivery platforms for OP than as therapeutic approaches ready for clinical application.195,210 Regulatory uncertainty remains an important external barrier to the commercialization of EV products. Dedicated regulatory pathways for EV products have not yet been fully harmonized, and current research is largely conducted in accordance with academic consensus guidelines such as MISEV2023.39 Naturally derived therapeutic EVs may be regulated as biological products, whereas engineered EVs loaded with genes or drugs may be classified as advanced therapy medicinal products or biological products depending on their active components and product characteristics. Their specific regulatory pathways therefore remain uncertain.211 A Trial-Design Framework for Clinical Validation Human studies of EV-based therapy remain at an early exploratory stage. Research in musculoskeletal disorders has mainly focused on knee osteoarthritis, cartilage repair, and localized tissue injury and therefore cannot be regarded as direct therapeutic evidence for systemic OP.4,53,203,212 No eligible OP-related T3 evidence was identified in the prespecified bibliographic databases searched in this review. Building on the product-standardization requirements outlined above, clinical protocols should translate validated product attributes into prespecified dosing, administration, batch-comparability, and safety procedures. Trial protocols and results should be reported in accordance with standards such as Standard Protocol Items: Recommendations for Interventional Trials (SPIRIT) and Consolidated Standards of Reporting Trials (CONSORT) to improve transparency, protocol reproducibility, and completeness of adverse-event reporting.213,214 Initial human studies in OP should not recruit broadly defined populations with highly heterogeneous etiologies. Priority should instead be given to participants with clearly defined clinical characteristics, high risk, and relatively homogeneous disease backgrounds, such as individuals with PMOP, GIOP, or another etiologically well-defined subtype.32,172,190 Participant selection should consider BMD, previous fragility fractures, bone-turnover status, age, renal function, inflammatory and metabolic status, and previous anti-OP treatment. These factors may influence both baseline fracture risk and the in vivo clearance, tissue exposure, and treatment response of EVs.198,202,215 Calcium and vitamin D supplementation, background anti-OP medication, and other concomitant therapies affecting bone metabolism should be prespecified and kept as consistent as possible. Early studies should position EV products as adjunctive interventions added to standard therapy rather than as replacements for established anti-OP treatment.28,198 Human studies may use a staged dose-escalation design, beginning with single-dose administration and proceeding to repeated administration according to safety findings and preliminary pharmacodynamic signals. The primary objectives of early-phase trials should be to establish safety, tolerability, an acceptable dose range, and the feasibility of administration rather than to directly demonstrate a reduction in fracture risk.216,217 For engineered EVs, early clinical studies should additionally evaluate the in vivo integrity of engineered modifications or loaded cargoes and monitor cargo detachment, off-target exposure, unintended biological effects, and immunological responses. Where technically feasible, early trials may incorporate exploratory pharmacokinetic and pharmacodynamic analyses. Early efficacy evaluation should use a time-structured hierarchy of endpoints. Short-term pharmacodynamic endpoints may include bone-turnover markers such as PINP and β-CTX to determine whether bone formation and bone resorption change in the anticipated directions.199 Intermediate endpoints may include lumbar-spine and total-hip BMD measured by DXA and, where feasible, changes in bone microarchitecture assessed using HR-pQCT or other imaging methods.5,198,199 Existing OP drug trials suggest that treatment-related changes in total-hip BMD are associated to some extent with reductions in fracture risk. Change in total-hip BMD at 24 months may therefore serve as an important intermediate endpoint but cannot replace evaluation of fracture events and long-term clinical benefit.5,201 In confirmatory studies, candidate EV products must demonstrate clinical benefits extending beyond short-term changes in bone turnover or BMD. Fragility fracture remains the most important clinical outcome of OP treatment, but assessment of fracture events requires large sample sizes and prolonged follow-up. A staged development pathway may therefore be adopted: early studies should establish safety, dose, and preliminary bone-metabolism signals; subsequent studies should evaluate changes in BMD and bone microarchitecture; and long-term follow-up should ultimately assess vertebral, hip, and other major fragility fractures. Long-term studies of conventional anti-OP drugs have shown that both therapeutic benefits and rare risks may change over time. EV products will likewise require long-term pharmacovigilance, particularly for immunological risks after repeated administration, abnormal bone remodeling, effects in ectopic tissues, and other delayed adverse events.218,219 Overall, clinical validation should proceed sequentially from early safety, tolerability, and dose exploration to assessment of pharmacodynamic activity, skeletal imaging outcomes, and ultimately fracture-related clinical benefit and long-term safety. Progression to larger confirmatory trials should depend on acceptable safety, a defined exposure–response relationship, and clinically meaningful skeletal signals, while the underlying product is required to satisfy the standardization criteria described above.186,195 Conclusion This scoping review indicates that non-bone-derived EVs and EV-like particles should be understood as a purpose-stratified evidence landscape rather than as a single, homogeneous biological mechanism. The available evidence remains dominated by cellular and animal studies, whereas rigorous in vivo source-to-bone attribution, participant-level validation, and evidence of clinical or industrial translation remain limited. The principal methodological contribution of this review lies in integrating purpose-based categories of mechanistic, biomarker, and intervention and delivery evidence with independent assessments of evidence strength and translational stage. Within this framework, Level C evidence is further differentiated into EV dependence, cargo attribution, and in vivo endogenous source attribution, thereby avoiding the interpretation of distinct research contexts at the same causal and translational level. Future research should complete the relevant evidence chains according to the intended translational objective. Mechanistic studies need to strengthen endogenous source tracing, validation of EV dependence, and causal cargo attribution. Biomarker studies require longitudinal cohorts and independent external validation and should assess whether candidate markers provide incremental value beyond BMD, bone-turnover markers, and existing risk-assessment tools.4,5,30,39,202 Intervention and delivery studies should quantitatively evaluate biodistribution, bone-tissue exposure, dose–response relationships, mechanism-related potency, and long-term safety, while establishing batch-consistent product systems compatible with GMP requirements.39,172 Until these issues are resolved, non-bone-derived EVs are more appropriately regarded as a candidate framework for mechanistic research, biomarker discovery, and nanomedicine development in OP than as clinically ready diagnostic or therapeutic platforms. This review also has several limitations. As a scoping review, its primary purpose was to map the structure of the evidence and identify research gaps; it cannot replace quantitative effect synthesis, formal risk-of-bias assessment, or evaluation of clinical efficacy.54,55 In addition, the available evidence showed marked geographical concentration. Although this distribution does not invalidate the findings of individual studies, it may limit the external generalizability of the evidence across research populations, experimental platforms, and healthcare settings. Gray literature, conference abstracts, and preprints were not systematically searched, and no formal assessment of publication bias was performed. The included studies also showed substantial heterogeneity in EV source, particle preparation and characterization methods, experimental models, dose units, and outcome measures. The conclusions therefore require further validation in standardized research systems and real-world patient populations.