Circulating miRNAs for glioblastoma monitoring: from biofluid to clinical decision Authors: Arthur Leclerc, Evelyne Emery, Guénaëlle Levallet Published: 2026 ABSTRACT Gliomas, particularly glioblastoma (GBM), remain highly lethal brain tumors with limited treatments. MRI-based monitoring lacks specificity in distinguishing progression from treatment effects. Circulating microRNAs (miRNAs), small non-coding RNAs involved in tumor biology, show promise as liquid biopsy biomarkers. We searched PubMed, Embase, Scopus, and Web of Science for clinical studies investigating circulating miRNAs as diagnostic, prognostic, or predictive biomarkers in glioma, across plasma, serum, and cerebrospinal fluid (CSF). Oncogenic miRNAs (oncomiRs) such as miR-21, miR-10b, miR-182, and miR-210 were consistently upregulated in glioma patients, whereas tumor suppressors including miR-124, miR-128, miR-137, and miR-181 family members were downregulated. Prognostically, elevated miR-21, miR-222, and miR-196b correlated with poor survival, while reduced miR-181d, miR-100, and miR-145 predicted worse outcomes. Composite signatures, particularly exosomal or CSF-derived panels, outperformed single markers. Dynamic changes in miR-21, miR-222, and miR-210 reflected treatment response or progression, with exosomal miR-1238 emerging as a marker of temozolomide resistance. Longitudinal serum levels of miR-223 and miR-320e correlated with MRI tumor volume and distinguished true progression from pseudo-progression. Circulating miRNAs show strong potential as minimally invasive biomarkers for glioma management, but clinical translation remains limited by technical and biological variability and lack of validation. Advances in biosensors, multi-omics, and standardization are key to clinical integration. Introduction Gliomas, the most prevalent primary brain tumors, pose significant clinical challenges due to their infiltrative nature, histological heterogeneity, and propensity for recurrence.1 Their classification, based on histopathological and molecular features, guides treatment decisions; however, the dynamic nature of these tumors often leads to therapeutic resistance and disease progression.2 Current clinical management relies heavily on magnetic resonance imaging (MRI) for tumor surveillance, yet MRI has well-recognized limitations in differentiating treatment-related changes from true tumor progression, often leading to ambiguous interpretations and the need for invasive procedures.3 Histological analysis, the cornerstone of glioma diagnosis, requires neurosurgery to obtain tumor tissue, presenting risks and discomfort to patients while failing to capture the full spatial and temporal heterogeneity of the tumor.4 These limitations underscore the urgent need for non-invasive, reliable biomarkers that can accurately reflect tumor status, predict treatment response, and detect early signs of recurrence.5 The concept of liquid biopsy originated from the description by Mandel and Métais in 1948 of cell-free nucleic acids in human plasma,6 but its translational potential emerged only decades later with the detection of tumor-derived DNA in the bloodstream of cancer patients.7 Early clinical applications focused on hematological malignancies, where peripheral blood directly reflects tumor burden and allows sampling of malignant clones for diagnosis and minimal residual disease assessment.8 Over the past two decades, the approach has expanded to solid organ malignancies, with circulating tumor DNA, circulating tumor cells, and extracellular vesicles becoming clinically actionable analytes in breast, lung, colorectal, and prostate cancers.9 The application of liquid biopsy to central nervous system disorders is more recent and faces specific challenges, notably the restricted exchange of biomolecules across the blood-brain barrier (BBB). Nevertheless, the use of cerebrospinal fluid (CSF) as a tumor-proximal compartment, together with growing interest in liquid biopsy for non-neoplastic CNS conditions such as neurodegenerative diseases and traumatic brain injury,10 has progressively positioned the field for clinical translation in neuro-oncology. Liquid biopsy—the analysis of circulating tumor-derived material in easily accessible biofluids—holds immense promise in neuro-oncology by offering a minimally invasive approach to monitor disease dynamics and personalize treatment strategies.11 It enables the detection of various tumor-derived components, including circulating tumor cells, cell-free DNA, circulating microRNAs (miRNAs), extracellular vesicles (EVs), proteins, and tumor-educated platelets.12 Among these, circulating miRNAs have emerged as particularly promising biomarkers due to their stability in biofluids, cancer-specific expression patterns, and ease of detection. Circulating miRNAs are small non-coding RNAs of approximately 18–22 nucleotides that regulate gene expression post-transcriptionally and exhibit remarkable stability in plasma, serum, CSF, and exosomes, owing to protection by RNA-binding proteins or encapsulation within vesicles.13 The dysregulation of specific miRNAs has been implicated in glioma initiation, progression, and treatment resistance, supporting their potential as biomarkers reflecting tumor biology in real time.14 This review assesses current research on circulating miRNAs as tools for clinical monitoring in glioma patients, covering their potential use in diagnosis, prognosis, treatment response assessment, and differentiation of recurrence from radiation necrosis. It also provides a comprehensive overview of circulating miRNA biology, detection methods and a systematic analysis of clinically relevant miRNA candidates, highlighting the challenges and opportunities in translating these biomarkers into clinical practice. Results Biological and technical background miRNAs: From molecular function to cancer biology miRNAs are evolutionarily conserved small non-coding RNAs of approximately 18–22 nucleotides that act as central post-transcriptional regulators of gene expression by guiding the RNA-induced silencing complex to partially complementary sequences in the 3′ untranslated regions of target messenger RNAs, leading to either translational repression or transcript degradation.15,16 Each miRNA can regulate hundreds of distinct targets, and most protein-coding genes are themselves subject to combinatorial control by multiple miRNAs, generating dense regulatory networks that govern fundamental processes including proliferation, differentiation, apoptosis, metabolism, and stress responses.17 Recent advances using structural biology, single-molecule imaging, and high-throughput functional screens have refined our understanding of the core biogenesis machinery, including the Microprocessor (Drosha-DGCR8) and Dicer complexes, and have revealed additional layers of regulation such as RNA editing, miRNA tailing, and target-directed miRNA decay.18 The relevance of miRNAs to oncology emerged from the seminal demonstration that the miR-15a/16-1 cluster is frequently deleted in chronic lymphocytic leukemia, establishing miRNAs as bona fide cancer genes.19 Subsequent large-scale profiling work showed that virtually all human cancers display characteristic miRNA expression patterns, often allowing accurate tumor classification by miRNA signatures alone.20 Functionally, miRNAs operate as oncogenes (oncogenic miRNAs [oncomiRs]) when their direct targets encode tumor suppressors and, conversely, as tumor suppressors when they silence oncogenic drivers.21 The miR-17-92 cluster typifies the context-dependent nature of these effects, behaving either as an oncogene or as a tumor suppressor depending on cellular background and tissue of origin.22 miR-21, one of the most extensively investigated oncomiRs, is upregulated across breast, lung, colorectal, prostate, and pancreatic cancers, where it represses pro-apoptotic targets such as PTEN and PDCD4 to promote proliferation, invasion, and resistance to therapy.23 Together with the let-7 family, miR-34a, miR-155, and the miR-200 family, these molecules collectively influence virtually all hallmarks of cancer. Beyond their intracellular regulatory function, miRNAs are released into the extracellular space within EVs, bound to RNA-binding proteins, or packaged in lipoprotein complexes, conferring remarkable stability in biofluids and positioning circulating miRNAs as attractive minimally invasive biomarkers across a broad spectrum of cancers.24 The translation of this biology toward clinical applications has progressed along two parallel tracks: as diagnostic and prognostic biomarkers in liquid biopsy approaches and as therapeutic targets through miRNA mimics, antagomirs, and modified oligonucleotides, several of which have advanced to early-phase clinical trials in oncology.25 Biology of miRNAs in glioma pathophysiology miRNA biogenesis is a multistep process beginning with transcription of primary miRNA transcripts (pri-miRNAs) by RNA polymerase II in the nucleus. These are processed by the Drosha-DGCR8 complex into precursor hairpins (pre-miRNAs), exported to the cytoplasm via Exportin-5, and cleaved by Dicer into ∼22-nucleotide duplexes. The guide strand is incorporated into the RNA-induced silencing complex (RISC), mainly bound to Argonaute proteins, enabling sequence-specific binding to the 3′ untranslated regions (UTRs) of target mRNAs and leading to translational repression or degradation.15,16 Gliomas exhibit extensive miRNA dysregulation, occurring through genetic alterations (e.g., copy number variations), epigenetic modifications (e.g., CpG island hypermethylation, histone deacetylation), and altered signaling pathways affecting miRNA processing.14,26,27,28,29 This results in a profound imbalance between oncomiRs and tumor-suppressor miRNAs, both contributing to tumor initiation and progression. Several miRNAs act as tumor suppressors. miR-7 and miR-128 are typically downregulated in glioblastoma (GBM), leading to unchecked activation of the EGFR/PI3K/AKT and E2F3a pathways, respectively, promoting proliferation and angiogenesis.30,31 Similarly, miR-34a induces apoptosis and suppresses migration but is consistently reduced in GBM tissue.32,33 Conversely, oncomiRs are upregulated: miR-21 promotes proliferation, invasion, and resistance to apoptosis by targeting PTEN, PDCD4, and TPM134,35,36; miR-10b facilitates invasion and angiogenesis through regulation of HOXD10, RhoC, and uPAR37; and miR-93 and miR-221/222 contribute to angiogenesis, cell cycle progression, and radiotherapy resistance.14,38 Altogether, miRNAs exert pleiotropic functions in GBM pathophysiology, affecting virtually all hallmarks of cancer. Their dynamic deregulation highlights both their potential as biomarkers for disease monitoring and as therapeutic targets.39,40 Circulation and stability in biofluids Circulating miRNAs are detectable in plasma, serum, and CSF.41,42 Their presence arises from active secretion via EVs—particularly exosomes and microvesicles—which protect them from enzymatic degradation43 and from passive release during apoptosis and necrosis.44 Stability is conferred by vesicular protection, association with RNA-binding proteins such as Argonaute-2, or encapsulation within lipoprotein complexes.45,46 These properties make circulating miRNAs consistently detectable in clinical specimens, as summarized in Table 1.Table 1Comparison of biofluids for circulating miRNA analysisAxisPlasmaSerumCSFKey referencesRelative miRNA load/abundancelower than serum in paired comparisons; sensitive to platelet depletion (prefer PFP)typically higher apparent levels than plasma, partly from platelet release during coagulationmuch lower RNA biomass overall; detectable but low abundanceWang, 201247; Sandau, 202448; Dufourd, 201949; Burgos, 201350Detected miRNA repertoire (NGS)not directly reported in these specific comparisons∼532 miRNAs detected (example dataset)∼486 miRNAs detected; 353 shared with serum; distinct profileBurgos, 201350Cross-matrix correlations–high inter-subject correlation (∼0.85)high inter-subject correlation (∼0.87); within-subject CSF–blood median Spearman ≈0.60Burgos, 201350Small-RNA class compositionhigh miRNA-to-tDR ratio among biofluids (e.g., ∼72)similar classes to plasma but shifted by coagulation effectscomposition differs from blood; lower mapping rates commonGodoy, 201851; Burgos, 201350; Alsop, 202252Longitudinal stability (healthy)many miRNAs stable over months under standardized PFP workflows; technical steps can dominate variance if not controlledexpected stable if clotting is standardized, but more variable than plasma due to coagulationnot directly quantified; detectability highly protocol-dependentSandau, 202448; Max, 201853; Wang, 201247; Burgos, 201350Typical input volumes for profiling200–500 μL commonly sufficient (qPCR/NGS) with PFP200–500 μL commonly used; document clotting conditions≥1 mL typically needed for robust detection; low-input methods essentialSandau, 202448; Dufourd, 201949; Burgos, 201350Pre-analytics to prioritizedouble/triple centrifugation to obtain PFP; consistent venipuncture and handlingstandardize clotting time/conditions; minimize platelet activationavoid blood contamination; swift processing; concentrate RNASandau, 202448; Wang, 201247; Burgos, 201350Major confoundersresidual platelets; variable extraction/inhibition if workflows differcoagulation/platelet release; platform biases across studiesextremely low input leading to adapter dimer and contaminant amplification; any blood carryoverFreedman, 201654; Wang, 201247; Alsop, 202252; Burgos, 201350Key characteristics of plasma, serum, and CSF are compared across miRNA abundance, detection repertoire, longitudinal stability, and pre-analytical requirements. CSF, cerebrospinal fluid; miRNA, microRNA; NGS, next-generation sequencing; PFP, platelet-free plasma; qPCR, quantitative polymerase chain reaction; RNA, ribonucleic acid; tDR, transfer RNA-derived fragment. Table 1Comparison of biofluids for circulating miRNA analysisAxisPlasmaSerumCSFKey referencesRelative miRNA load/abundancelower than serum in paired comparisons; sensitive to platelet depletion (prefer PFP)typically higher apparent levels than plasma, partly from platelet release during coagulationmuch lower RNA biomass overall; detectable but low abundanceWang, 201247; Sandau, 202448; Dufourd, 201949; Burgos, 201350Detected miRNA repertoire (NGS)not directly reported in these specific comparisons∼532 miRNAs detected (example dataset)∼486 miRNAs detected; 353 shared with serum; distinct profileBurgos, 201350Cross-matrix correlations–high inter-subject correlation (∼0.85)high inter-subject correlation (∼0.87); within-subject CSF–blood median Spearman ≈0.60Burgos, 201350Small-RNA class compositionhigh miRNA-to-tDR ratio among biofluids (e.g., ∼72)similar classes to plasma but shifted by coagulation effectscomposition differs from blood; lower mapping rates commonGodoy, 201851; Burgos, 201350; Alsop, 202252Longitudinal stability (healthy)many miRNAs stable over months under standardized PFP workflows; technical steps can dominate variance if not controlledexpected stable if clotting is standardized, but more variable than plasma due to coagulationnot directly quantified; detectability highly protocol-dependentSandau, 202448; Max, 201853; Wang, 201247; Burgos, 201350Typical input volumes for profiling200–500 μL commonly sufficient (qPCR/NGS) with PFP200–500 μL commonly used; document clotting conditions≥1 mL typically needed for robust detection; low-input methods essentialSandau, 202448; Dufourd, 201949; Burgos, 201350Pre-analytics to prioritizedouble/triple centrifugation to obtain PFP; consistent venipuncture and handlingstandardize clotting time/conditions; minimize platelet activationavoid blood contamination; swift processing; concentrate RNASandau, 202448; Wang, 201247; Burgos, 201350Major confoundersresidual platelets; variable extraction/inhibition if workflows differcoagulation/platelet release; platform biases across studiesextremely low input leading to adapter dimer and contaminant amplification; any blood carryoverFreedman, 201654; Wang, 201247; Alsop, 202252; Burgos, 201350Key characteristics of plasma, serum, and CSF are compared across miRNA abundance, detection repertoire, longitudinal stability, and pre-analytical requirements. CSF, cerebrospinal fluid; miRNA, microRNA; NGS, next-generation sequencing; PFP, platelet-free plasma; qPCR, quantitative polymerase chain reaction; RNA, ribonucleic acid; tDR, transfer RNA-derived fragment. Comparison of biofluids for circulating miRNA analysis Key characteristics of plasma, serum, and CSF are compared across miRNA abundance, detection repertoire, longitudinal stability, and pre-analytical requirements. CSF, cerebrospinal fluid; miRNA, microRNA; NGS, next-generation sequencing; PFP, platelet-free plasma; qPCR, quantitative polymerase chain reaction; RNA, ribonucleic acid; tDR, transfer RNA-derived fragment. When comparing biofluids, three patterns emerge: serum typically shows higher apparent miRNA levels than plasma due to platelet activation during coagulation47,49,53; CSF has lower RNA biomass yet hundreds of miRNAs remain detectable, with only moderate concordance between CSF and blood profiles,50 and small-RNA composition differs across biofluids and between bulk versus EV-enriched fractions.51,55,56,57 Longitudinal work shows that, under standardized platelet-free plasma workflows, many miRNAs are highly stable over months, though technical variability dominates if extraction is not tightly controlled.48,53,54 Practically, platelet-free plasma is the most reproducible choice, serum is usable if coagulation is standardized, and CSF demands larger input volumes and low-input-optimized protocols.50,52,53,58 Pre-analytical variability Pre-analytical variables—sample collection, processing, storage, and RNA extraction—significantly impact the accuracy and reproducibility of circulating miRNA measurements.59 The choice of anticoagulant, venipuncture technique, and centrifugation speed directly affect plasma and serum composition.60 Circulating miRNAs remain stable in whole blood for up 48 h. Key recommendations for miRNA quality preservation include use of EDTA-containing collection tubes with processing within 4 h,61 storage at −80°C,61 minimization of freeze-thaw cycles,61,62 use of extraction kits specifically designed for small RNAs,61 and harmonization of all pre-analytical steps including the use of exogenous spike-ins.61 Specialized stabilization tubes (e.g., Streck Nucleic Acid BCT) can maintain cell-free RNA/miRNA concentrations for up to 7 days at ambient temperature.63 Contamination with white blood cells from inflammatory processes should be considered during interpretation.64 Detection methods Several platforms are available for circulating miRNA quantification, each with distinct advantages (Table 2). Real-time RT-qPCR offers high sensitivity, specificity, and dynamic range for absolute quantification of small miRNA panels.66,67 Microarray platforms enable high-throughput profiling of hundreds to thousands of miRNAs simultaneously using immobilized complementary probes.68,69 Next-generation sequencing (NGS; small RNA sequencing [RNA-seq]) provides comprehensive, unbiased detection of both known and novel miRNAs.70,71 The selection of detection platform should align with the study objective: RT-qPCR for focused validation, microarrays for large-scale discovery, and NGS for comprehensive profiling.Table 2Analytical platforms for circulating miRNA detection: principles, advantages, and limitationsMethodPrincipleAdvantagesLimitationsBest use case/key referencesRT-qPCRreverse transcription of mature miRNAs into cDNA, followed by PCR amplification with specific primers (SYBR Green or TaqMan probes)high sensitivity and specificity; quantitative (absolute or relative); relatively low cost and fast turnaroundlimited to known miRNAs; susceptible to normalization bias; small panels onlyvalidation of candidate miRNAs in clinical studies (Chen, 202136; Kroh, 201065)Microarrayhybridization of fluorescently labeled RNA to immobilized probes complementary to known miRNAshigh-throughput (hundreds to thousands of miRNAs); relatively low cost per sample; well established protocolslower sensitivity than RT-qPCR or NGS; Limited to annotated miRNAs; Semi-quantitativediscovery of differential expression profiles in large cohorts (Godoy, 201851; Wang, 2017)NGS (small RNA-seq)deep sequencing of barcoded small-RNA libraries generated from biofluidscomprehensive (detects both known and novel miRNAs); high resolution (isomiRs, isoforms); quantitative and unbiasedexpensive and time-consuming; requires advanced bioinformatics; sensitive to library prep biasdiscovery of new miRNAs, isoform analysis, complete profiling (O’Brien, 201816)cDNA, complementary DNA; miRNA, microRNA; NGS, next-generation sequencing; PCR, polymerase chain reaction; RT-qPCR, quantitative reverse transcription PCR; RNA, ribonucleic acid; RNA-seq, RNA sequencing; SYBR, SYBR Green fluorescent dye. Table 2Analytical platforms for circulating miRNA detection: principles, advantages, and limitationsMethodPrincipleAdvantagesLimitationsBest use case/key referencesRT-qPCRreverse transcription of mature miRNAs into cDNA, followed by PCR amplification with specific primers (SYBR Green or TaqMan probes)high sensitivity and specificity; quantitative (absolute or relative); relatively low cost and fast turnaroundlimited to known miRNAs; susceptible to normalization bias; small panels onlyvalidation of candidate miRNAs in clinical studies (Chen, 202136; Kroh, 201065)Microarrayhybridization of fluorescently labeled RNA to immobilized probes complementary to known miRNAshigh-throughput (hundreds to thousands of miRNAs); relatively low cost per sample; well established protocolslower sensitivity than RT-qPCR or NGS; Limited to annotated miRNAs; Semi-quantitativediscovery of differential expression profiles in large cohorts (Godoy, 201851; Wang, 2017)NGS (small RNA-seq)deep sequencing of barcoded small-RNA libraries generated from biofluidscomprehensive (detects both known and novel miRNAs); high resolution (isomiRs, isoforms); quantitative and unbiasedexpensive and time-consuming; requires advanced bioinformatics; sensitive to library prep biasdiscovery of new miRNAs, isoform analysis, complete profiling (O’Brien, 201816)cDNA, complementary DNA; miRNA, microRNA; NGS, next-generation sequencing; PCR, polymerase chain reaction; RT-qPCR, quantitative reverse transcription PCR; RNA, ribonucleic acid; RNA-seq, RNA sequencing; SYBR, SYBR Green fluorescent dye. Analytical platforms for circulating miRNA detection: principles, advantages, and limitations cDNA, complementary DNA; miRNA, microRNA; NGS, next-generation sequencing; PCR, polymerase chain reaction; RT-qPCR, quantitative reverse transcription PCR; RNA, ribonucleic acid; RNA-seq, RNA sequencing; SYBR, SYBR Green fluorescent dye. RT-qPCR RT-qPCR remains the most widely used method for focused validation of circulating miRNA biomarkers, offering high sensitivity, specificity, and dynamic range for absolute quantification of small miRNA panels.66,67 It has been the technique of choice for clinical translation studies in solid malignancies, as illustrated by a recent multicenter European low-dose CT screening program in which a nine-miRNA plasma signature was deployed by RT-qPCR for early detection of lung cancer with consistent diagnostic accuracy across cohorts.72 Its main limitations include restricted multiplexing capacity and a dependence on prior sequence knowledge. Microarrays Microarray platforms enable high-throughput profiling of hundreds to thousands of miRNAs simultaneously using immobilized complementary probes, providing semi-quantitative expression maps useful for hypothesis-generating studies.68,69 They have been extensively applied to identify miRNA signatures in breast, lung, and colorectal cancers but are gradually being superseded by sequencing-based approaches given concerns regarding cross-hybridization and reduced sensitivity for low-abundance miRNAs.73 NGS (small RNA-seq) NGS of small RNAs provides comprehensive, unbiased detection of both known and novel miRNAs, including isomiRs, sequence variants, and other small non-coding RNA species.70,71 Small RNA-seq has been instrumental in characterizing the complete circulating miRNA landscape across biofluids and has been widely applied across solid malignancies for biomarker discovery; however, the higher cost, the requirement for specialized bioinformatic pipelines, and the need for careful normalization currently limit its routine clinical use. Droplet digital PCR Droplet digital PCR (ddPCR) represents an emerging quantitative platform offering absolute measurement of low-abundance circulating miRNAs without reliance on a calibration curve, with high precision and improved tolerance to inhibitors commonly present in plasma and serum.74 ddPCR has been successfully applied to circulating miRNA quantification in ovarian, breast, and lung cancers and is increasingly attractive for monitoring minimal residual disease and dynamic biomarker variations during treatment. Electrochemical biosensors and emerging platforms A range of biosensor formats—including electrochemical, plasmonic, and microfluidic devices—is under active development to address the unmet need for rapid, low-cost, and point-of-care miRNA detection. Recent electrochemical biosensors enhanced with nanomaterials have demonstrated highly sensitive and specific detection of disease-associated miRNAs in liver and pancreatic cancers, with attomolar detection limits reported in some configurations.75,76 Although their clinical implementation remains limited, these platforms are particularly promising for serial monitoring scenarios such as post-operative surveillance in neuro-oncology. Clinical applications of circulating miRNAs in glioma Diagnostic biomarkers Identifying reliable biomarkers for early GBM diagnosis remains challenging given the limited specificity of conventional neuroimaging. Numerous studies have evaluated the diagnostic potential of circulating miRNAs in plasma, serum, or CSF. The main candidates are summarized in Table 3.Table 3Circulating miRNAs with diagnostic potential in glioma patientsmiRFluidExpression vs. healthyCellular function (succinct)Authors (name, year)miR-1plasma/serumvariableinvasion (TKS4/TKS5/EFHD2)Herman, 201577; Díaz Méndez, 202378let-7aCSFincreasedtumor suppressor (context dependent in CSF panel)Kopková, 201941let-7bCSFincreasedtumor suppressor (context-dependent in CSF panel)Kopková, 201941miR-10aCSFincreasedstemness/invasionKopková, 201941; Teplyuk, 201279miR-10bCSFincreasedstemness/invasionTeplyuk, 201279; Kopková, 201941miR-15b-3pserum (CD44+ EVs)increasedno established function in glioma (diagnostic biomarker only)Tzaridis, 202080miR-21CSF (exosomes/EVs); serum/plasmaincreasedOncomiR; targets PTEN/PDCD4; invasionAkers, 201381; Baraniskin, 201282; Teplyuk, 201279; Kopková, 201941; Wang, 201283; Parvizhamidi, 201884; Tzaridis, 202080miR-21-3pserum (CD44+ EVs)increasedOncomiRTzaridis, 202080miR-23a-3pplasmaincreasedoncogene; invasion via HOXD10, PI3K/AKTHerman, 201577; Hu, 201385; Yachi, 201886miR-26aserumincreasedoncogene; targets PTENParvizhamidi, 201884; Huse, 200987miR-29bserum (exosomes)decreasedtumor suppressor; ECM regulationZhong, 201988miR-29a-3pplasmaincreasedproliferation suppressorGéczi, 202189miR-30eCSFincreasedcell cycle/apoptosisKopková, 201941miR-34aserumdecreasedtumor suppressor; EGFR axisRahmati, 202190miR-34b-3pplasma (exosomes)increasedtumor suppressorYamashita, 202191miR-100serumdecreasedsuppresses proliferation/migrationZhang, 201992miR-106a-5pserum (CD44+ EVs)variablesuppressor; targets SLC2A3, E2F1, FASTKTzaridis, 202080;miR-124plasma/CSFdecreasedtumor suppressor; neuronal differentiationHerman, 201577; Akers, 201793miR-128plasma/CSFdecreasedtumor suppressor; neuronal identityHerman, 201577; Akers, 201793miR-137serumdecreasedtumor suppressor; differentiationLi, 201694miR-140CSFincreasedmigration/ECM regulationKopková, 201941miR-181dplasma/serum (exosomes)decreasedpredictive of MGMT; treatment responseSippl, 202295; Stakaitis, 202096miR-181aplasmadecreasedtumor suppressorWu, 202297miR-181bplasmadecreasedtumor suppressorWu, 202297miR-182plasmaincreasedoncogene; targets FOXO3, BCL2Xiao, 201698; Kouri, 201599miR-195-5pplasmaincreasedcell cycle regulationGéczi, 202189miR-196aCSFincreasedoncogene; HOX gene regulationKopková, 201941miR-196bCSFincreasedoncogene; poor prognosisKopková, 201941miR-210serum; serum exosomesincreasedhypoxia oncomiRLai, 2015100; Lan, 2020101miR-222CSF; serum (exosomes)increasedoncogene; invasion; DNA repairAkers, 201793; Olioso, 2021102miR-302c-3pplasmaincreasedsuppressor; inhibits proliferation/invasion (targets MTDH)Herman, 201577miR-320serum (exosomes)increasedsuppressor; inhibits SND1/β-cateninManterola, 2014103; Li, 2017104miR-342-3pplasmaDysregulatedneuronal identityWang, 201283miR-363serumincreaseddiagnostic biomarkerVojdani, 2021105miR-383plasmadecreasedsuppressor; proliferation/invasionHerman, 201577miR-433-3pplasmaincreasedneuronal developmentGéczi, 202189miR-451aserumchangedmetabolic rheostat; LKB1/AMPK via CAB39Godlewski, 2010106; Kim, 2011107miR-485-3pserumdecreasedsuppressor; inhibits RNF135Wang, 2017108miR-485-5pserumdecreasedtumor-suppressor biomarkerRahmati, 202190miR-574-3pserum (microvesicles)increasedbiomarker; mechanistic role unclearManterola, 2014103; Jiang, 2018109; Peng, 2022110miR-582-5pserumincreaseddiagnostic biomarkerVojdani, 2021105miR-711CSFincreaseddifferential vs. medulloblastomaYang, 2013111miR-935CSFdecreaseddifferential vs. medulloblastomaYang, 2013111miR-2276-5pplasma (exosomes)decreasedtargets RAB13Sun, 2021112miR-4298serumchangedno function reported in gliomaWang, 2017108Expression changes are reported relative to healthy controls. AKT, protein kinase B; AMPK, AMP-activated protein kinase; BCL2, B cell lymphoma 2; CAB39, calcium-binding protein 39; CD44, cluster of differentiation 44; CSF, cerebrospinal fluid; ECM, extracellular matrix; EFHD2, EF-hand domain family member D2; EGFR, epidermal growth factor receptor; EV, extracellular vesicle; FASTK, Fas-activated serine/threonine kinase; FOXO3, forkhead box O3; HOX, homeobox; HOXD10, homeobox D10; LKB1, liver kinase B1; MGMT, O6-methylguanine-DNA methyltransferase; miR, microRNA; MTDH, metadherin; PDCD4, programmed cell death 4; PI3K, phosphoinositide 3-kinase; PTEN, phosphatase and tensin homolog; RAB13, Ras-related protein Rab-13; RNF135, ring finger protein 135; SLC2A3, solute carrier family 2 member 3; SND1, staphylococcal nuclease and Tudor domain containing 1; TKS4/TKS5, tyrosine kinase substrate 4/5. Table 3Circulating miRNAs with diagnostic potential in glioma patientsmiRFluidExpression vs. healthyCellular function (succinct)Authors (name, year)miR-1plasma/serumvariableinvasion (TKS4/TKS5/EFHD2)Herman, 201577; Díaz Méndez, 202378let-7aCSFincreasedtumor suppressor (context dependent in CSF panel)Kopková, 201941let-7bCSFincreasedtumor suppressor (context-dependent in CSF panel)Kopková, 201941miR-10aCSFincreasedstemness/invasionKopková, 201941; Teplyuk, 201279miR-10bCSFincreasedstemness/invasionTeplyuk, 201279; Kopková, 201941miR-15b-3pserum (CD44+ EVs)increasedno established function in glioma (diagnostic biomarker only)Tzaridis, 202080miR-21CSF (exosomes/EVs); serum/plasmaincreasedOncomiR; targets PTEN/PDCD4; invasionAkers, 201381; Baraniskin, 201282; Teplyuk, 201279; Kopková, 201941; Wang, 201283; Parvizhamidi, 201884; Tzaridis, 202080miR-21-3pserum (CD44+ EVs)increasedOncomiRTzaridis, 202080miR-23a-3pplasmaincreasedoncogene; invasion via HOXD10, PI3K/AKTHerman, 201577; Hu, 201385; Yachi, 201886miR-26aserumincreasedoncogene; targets PTENParvizhamidi, 201884; Huse, 200987miR-29bserum (exosomes)decreasedtumor suppressor; ECM regulationZhong, 201988miR-29a-3pplasmaincreasedproliferation suppressorGéczi, 202189miR-30eCSFincreasedcell cycle/apoptosisKopková, 201941miR-34aserumdecreasedtumor suppressor; EGFR axisRahmati, 202190miR-34b-3pplasma (exosomes)increasedtumor suppressorYamashita, 202191miR-100serumdecreasedsuppresses proliferation/migrationZhang, 201992miR-106a-5pserum (CD44+ EVs)variablesuppressor; targets SLC2A3, E2F1, FASTKTzaridis, 202080;miR-124plasma/CSFdecreasedtumor suppressor; neuronal differentiationHerman, 201577; Akers, 201793miR-128plasma/CSFdecreasedtumor suppressor; neuronal identityHerman, 201577; Akers, 201793miR-137serumdecreasedtumor suppressor; differentiationLi, 201694miR-140CSFincreasedmigration/ECM regulationKopková, 201941miR-181dplasma/serum (exosomes)decreasedpredictive of MGMT; treatment responseSippl, 202295; Stakaitis, 202096miR-181aplasmadecreasedtumor suppressorWu, 202297miR-181bplasmadecreasedtumor suppressorWu, 202297miR-182plasmaincreasedoncogene; targets FOXO3, BCL2Xiao, 201698; Kouri, 201599miR-195-5pplasmaincreasedcell cycle regulationGéczi, 202189miR-196aCSFincreasedoncogene; HOX gene regulationKopková, 201941miR-196bCSFincreasedoncogene; poor prognosisKopková, 201941miR-210serum; serum exosomesincreasedhypoxia oncomiRLai, 2015100; Lan, 2020101miR-222CSF; serum (exosomes)increasedoncogene; invasion; DNA repairAkers, 201793; Olioso, 2021102miR-302c-3pplasmaincreasedsuppressor; inhibits proliferation/invasion (targets MTDH)Herman, 201577miR-320serum (exosomes)increasedsuppressor; inhibits SND1/β-cateninManterola, 2014103; Li, 2017104miR-342-3pplasmaDysregulatedneuronal identityWang, 201283miR-363serumincreaseddiagnostic biomarkerVojdani, 2021105miR-383plasmadecreasedsuppressor; proliferation/invasionHerman, 201577miR-433-3pplasmaincreasedneuronal developmentGéczi, 202189miR-451aserumchangedmetabolic rheostat; LKB1/AMPK via CAB39Godlewski, 2010106; Kim, 2011107miR-485-3pserumdecreasedsuppressor; inhibits RNF135Wang, 2017108miR-485-5pserumdecreasedtumor-suppressor biomarkerRahmati, 202190miR-574-3pserum (microvesicles)increasedbiomarker; mechanistic role unclearManterola, 2014103; Jiang, 2018109; Peng, 2022110miR-582-5pserumincreaseddiagnostic biomarkerVojdani, 2021105miR-711CSFincreaseddifferential vs. medulloblastomaYang, 2013111miR-935CSFdecreaseddifferential vs. medulloblastomaYang, 2013111miR-2276-5pplasma (exosomes)decreasedtargets RAB13Sun, 2021112miR-4298serumchangedno function reported in gliomaWang, 2017108Expression changes are reported relative to healthy controls. AKT, protein kinase B; AMPK, AMP-activated protein kinase; BCL2, B cell lymphoma 2; CAB39, calcium-binding protein 39; CD44, cluster of differentiation 44; CSF, cerebrospinal fluid; ECM, extracellular matrix; EFHD2, EF-hand domain family member D2; EGFR, epidermal growth factor receptor; EV, extracellular vesicle; FASTK, Fas-activated serine/threonine kinase; FOXO3, forkhead box O3; HOX, homeobox; HOXD10, homeobox D10; LKB1, liver kinase B1; MGMT, O6-methylguanine-DNA methyltransferase; miR, microRNA; MTDH, metadherin; PDCD4, programmed cell death 4; PI3K, phosphoinositide 3-kinase; PTEN, phosphatase and tensin homolog; RAB13, Ras-related protein Rab-13; RNF135, ring finger protein 135; SLC2A3, solute carrier family 2 member 3; SND1, staphylococcal nuclease and Tudor domain containing 1; TKS4/TKS5, tyrosine kinase substrate 4/5. Circulating miRNAs with diagnostic potential in glioma patients Expression changes are reported relative to healthy controls. AKT, protein kinase B; AMPK, AMP-activated protein kinase; BCL2, B cell lymphoma 2; CAB39, calcium-binding protein 39; CD44, cluster of differentiation 44; CSF, cerebrospinal fluid; ECM, extracellular matrix; EFHD2, EF-hand domain family member D2; EGFR, epidermal growth factor receptor; EV, extracellular vesicle; FASTK, Fas-activated serine/threonine kinase; FOXO3, forkhead box O3; HOX, homeobox; HOXD10, homeobox D10; LKB1, liver kinase B1; MGMT, O6-methylguanine-DNA methyltransferase; miR, microRNA; MTDH, metadherin; PDCD4, programmed cell death 4; PI3K, phosphoinositide 3-kinase; PTEN, phosphatase and tensin homolog; RAB13, Ras-related protein Rab-13; RNF135, ring finger protein 135; SLC2A3, solute carrier family 2 member 3; SND1, staphylococcal nuclease and Tudor domain containing 1; TKS4/TKS5, tyrosine kinase substrate 4/5. miR-21 is the most extensively studied, consistently overexpressed in serum, plasma, and CSF exosomes of GBM patients. It acts as an oncomiR by repressing PTEN and PDCD4, promoting proliferation, invasion, and apoptosis resistance.41,79,80,81,82,83,84,95,113 Isoform-specific analyses, such as enrichment of miR-21-3p in CD44+ EVs, suggest that vesicle subtypes may carry more discriminative signatures than bulk specimens.113 Other upregulated oncomiRs include miR-10a/b, associated with stemness and invasion in CSF41,83; miR-210, a hypoxia-regulated miRNA reflecting tumor hypoxia and angiogenesis100,101; and miR-182, elevated in plasma and promoting proliferation via FOXO3 and BCL2.98,99 Additional upregulated candidates include miR-23a-3p, miR-26a, miR-30e, miR-140, miR-195-5p, miR-29a-3p, miR-433-3p, miR-363, miR-582-5p, miR-574-3p, and exosomal miR-320.41,77,80,85,86,87,89,103,104,105,109,110 Conversely, several tumor-suppressor miRNAs are consistently downregulated: miR-124 and miR-128 (neuronal differentiation-associated, decreased in both plasma and CSF),77,93 miR-137,94 miR-100,92 miR-34a/b (regulating EGFR and cell cycle checkpoints),90,91 as well as miR-383, miR-29b, miR-485-3p/5p, and miR-181 family members.77,81,88,96,108,114,115 Notably, CSF-derived miR-711 (upregulated) and miR-935 (downregulated) may differentiate GBM from medulloblastoma.111 Some biomarkers show context-dependent expression: miR-1 has been reported as both increased and decreased,77,78 miR-106a-5p appears enriched in CD44+ EVs but may function as a tumor suppressor,113 and miR-451a acts as a metabolic rheostat via the LKB1/AMPK/CAB39 pathway.106,107 A recurring limitation is that single biomarkers often lack reproducibility across cohorts. Composite signatures have consistently demonstrated superior diagnostic accuracy: CSF-based panels combining let-7a/b, miR-10a/b, miR-30e, miR-140, and miR-196a/b yielded excellent sensitivity and specificity41; EV-based signatures including miR-21, miR-21-3p, miR-15b-3p, and miR-106a-5p improved discrimination113; and plasma- and serum-based panels further confirmed the added value of multimarker approaches.89,105 These data suggest that miRNA signatures, rather than individual miRNAs, represent the most reliable strategy for non-invasive GBM diagnosis. Prognostic biomarkers Circulating miRNAs have been extensively investigated as prognostic biomarkers, with several candidates showing correlations with overall survival (OS) and progression-free survival (PFS) (Table 4).Table 4Circulating miRNAs with prognostic value in gliomamiRFluid/matrixVariation associated with prognosisPrognosisNotesAuthors (name, year)miR-1-3pserumlow → poor OS/PFSpoorpart of 3-miR serum signature (with miR-26a-1-3p, miR-487b-3p); lower levels stratify IDHwt and associate with worse survivalDíaz Méndez, 202378miR-10bCSF/serumhigh → poor OSpoorhigh CSF levels associate with median OS ∼9 vs. 16.5 months (low)Kopková, 201941miR-15b-3pserum EV (CD44+)high → poor outcomepoorincreased in EVs; linked to poor survivalTzaridis, 202080miR-20a-5pserumhigh → poor DFSpoorhigh levels associate with decreased survivalZhao, 2017116miR-21serum/CSF exosomes; EVshigh → poor OS/recurrencepoorOncomiR; elevated levels correlate with shorter OS and higher recurrence riskOlioso, 2021102; Shi, 2015117; Tzaridis, 202080; Joo Park, 2022113miR-21-3pserum EV (CD44+)high → poor outcomepoorisoform enriched in EVs; associated with adverse outcomeTzaridis, 202080miR-26a-1-3pserumlow → poor OS/PFSpoorwith miR-1-3p & miR-487b-3p in serum signature; lower = worseDíaz Méndez, 202378miR-29bserum (exosomes)low → poor OSpoorunderexpressed in GBM; low exosomal levels predict shorter survival; increase post-surgeryZhong, 201988miR-100serumlow → poor outcomepoorlower serum miR-100 predicts worse outcomeZhang, 2019b92miR-106a-5pserumhigh → poor OS/DFSpoorhigh levels associate with poorer outcomesZhao, 2017116miR-137serumlow → poor prognosispoordecreased in GBM; lower levels predict poorer prognosisLi, 201694miR-145-5pserumhigh → better OS/DFSBetterhigher serum miR-145-5p independently predicts improved survivalZhang, 2019a118; Zhao, 2017116miR-145-5pserumlow → poor OS/DFSpoorindependent prognostic indicator; lower level associates with worse survivalZhang, 2019a118; Zhao, 2017116miR-151aCSF exosomes/serumlow → poor OS/PFS (TMZ-treated)poorlower expression associated with worse survival; restoration sensitizes to TMZZeng, 2018119miR-155serum/plasma/CSFhigh → poor outcomepoorelevated levels linked to adverse prognosisBustos, 2022120; Ordoñez-Rubiano, 202440miR-181bplasmalow → poor prognosispoorlower concentrations associate with worse outcome; inverse relation with functional scoresWu, 202297; Stakaitis, 202096miR-181dplasma/serum; tumorlow → poor prognosispoorlower blood levels associate with worse prognosis; also predictive of MGMT/treatment responseSippl, 202295; Stakaitis, 202096miR-181aplasmalow → poor prognosispoorlower concentrations associate with worse outcomeWu, 202297miR-182serumhigh → poor OS/DFSpoorhigh pre-treatment serum levels associate with poor prognosisZhao, 2017116miR-185plasmalow → poor OSpoordecline in plasma level associated with poor survival; glioma-specific vs. other CNS tumorsTang, 2015121miR-193bserumhigh → poor OSpoorpromotes proliferation/invasion; high predicts poor survivalZhu, 2019122miR-196bCSF/serumhigh → poor OSpooroverexpressed; poor survival associationKopková, 201941miR-203serumlow → poor OS/PFSpoorreduced circulating miR-203 predicts worse OS/PFSChen, 2017b123miR-205serumlow → poor OS/PFSpoordecreased levels associated with worse survivalYue, 2016124miR-210serum/Exosomeshigh → poor OSpoorhypoxia-related; high levels predict poor survivalLan, 2020101miR-221serum/plasma/CSFhigh → poor outcomepoorelevated in liquid biopsies; associated with prognosisBustos, 2022120; Ordoñez-Rubiano, 202440miR-222serum (total/exosomal); CSFhigh → poor OS/DFSpoorOncomiR; higher levels predict decreased survivalOlioso, 2021102; Zhao, 2017116miR-328-3pserum EV (CD44+)high → poor outcomepoorincluded among increased EV-miRNAs linked to poor survivalTzaridis, 202080miR-485-3pserumlow → poor OSpoorlow serum level predicts shorter survivalWang, 2017108miR-487b-3pserumlow → poor OS/PFSpoorwith miR-1-3p & miR-26a-1-3p in serum signature; lower = worseDíaz Méndez, 202378miR-663serum/tissuelow → poor outcomepoorlower miR-663 associated with unfavorable clinical outcomeChen, 2017b123miR-2276-5pplasma (exosomes)low → poor OSpoortargets RAB13; low in exosomes associates with higher grade and shorter survivalSun, 2021112Associations with OS, PFS, and DFS are reported. CD44, cluster of differentiation 44; CNS, central nervous system; CSF, cerebrospinal fluid; DFS, disease-free survival; EV, extracellular vesicle; GBM, glioblastoma; IDHwt, isocitrate dehydrogenase wild-type; MGMT, O6-methylguanine-DNA methyltransferase; miR, microRNA; OS, overall survival; PFS, progression-free survival; RT, radiotherapy; TMZ, temozolomide. Table 4Circulating miRNAs with prognostic value in gliomamiRFluid/matrixVariation associated with prognosisPrognosisNotesAuthors (name, year)miR-1-3pserumlow → poor OS/PFSpoorpart of 3-miR serum signature (with miR-26a-1-3p, miR-487b-3p); lower levels stratify IDHwt and associate with worse survivalDíaz Méndez, 202378miR-10bCSF/serumhigh → poor OSpoorhigh CSF levels associate with median OS ∼9 vs. 16.5 months (low)Kopková, 201941miR-15b-3pserum EV (CD44+)high → poor outcomepoorincreased in EVs; linked to poor survivalTzaridis, 202080miR-20a-5pserumhigh → poor DFSpoorhigh levels associate with decreased survivalZhao, 2017116miR-21serum/CSF exosomes; EVshigh → poor OS/recurrencepoorOncomiR; elevated levels correlate with shorter OS and higher recurrence riskOlioso, 2021102; Shi, 2015117; Tzaridis, 202080; Joo Park, 2022113miR-21-3pserum EV (CD44+)high → poor outcomepoorisoform enriched in EVs; associated with adverse outcomeTzaridis, 202080miR-26a-1-3pserumlow → poor OS/PFSpoorwith miR-1-3p & miR-487b-3p in serum signature; lower = worseDíaz Méndez, 202378miR-29bserum (exosomes)low → poor OSpoorunderexpressed in GBM; low exosomal levels predict shorter survival; increase post-surgeryZhong, 201988miR-100serumlow → poor outcomepoorlower serum miR-100 predicts worse outcomeZhang, 2019b92miR-106a-5pserumhigh → poor OS/DFSpoorhigh levels associate with poorer outcomesZhao, 2017116miR-137serumlow → poor prognosispoordecreased in GBM; lower levels predict poorer prognosisLi, 201694miR-145-5pserumhigh → better OS/DFSBetterhigher serum miR-145-5p independently predicts improved survivalZhang, 2019a118; Zhao, 2017116miR-145-5pserumlow → poor OS/DFSpoorindependent prognostic indicator; lower level associates with worse survivalZhang, 2019a118; Zhao, 2017116miR-151aCSF exosomes/serumlow → poor OS/PFS (TMZ-treated)poorlower expression associated with worse survival; restoration sensitizes to TMZZeng, 2018119miR-155serum/plasma/CSFhigh → poor outcomepoorelevated levels linked to adverse prognosisBustos, 2022120; Ordoñez-Rubiano, 202440miR-181bplasmalow → poor prognosispoorlower concentrations associate with worse outcome; inverse relation with functional scoresWu, 202297; Stakaitis, 202096miR-181dplasma/serum; tumorlow → poor prognosispoorlower blood levels associate with worse prognosis; also predictive of MGMT/treatment responseSippl, 202295; Stakaitis, 202096miR-181aplasmalow → poor prognosispoorlower concentrations associate with worse outcomeWu, 202297miR-182serumhigh → poor OS/DFSpoorhigh pre-treatment serum levels associate with poor prognosisZhao, 2017116miR-185plasmalow → poor OSpoordecline in plasma level associated with poor survival; glioma-specific vs. other CNS tumorsTang, 2015121miR-193bserumhigh → poor OSpoorpromotes proliferation/invasion; high predicts poor survivalZhu, 2019122miR-196bCSF/serumhigh → poor OSpooroverexpressed; poor survival associationKopková, 201941miR-203serumlow → poor OS/PFSpoorreduced circulating miR-203 predicts worse OS/PFSChen, 2017b123miR-205serumlow → poor OS/PFSpoordecreased levels associated with worse survivalYue, 2016124miR-210serum/Exosomeshigh → poor OSpoorhypoxia-related; high levels predict poor survivalLan, 2020101miR-221serum/plasma/CSFhigh → poor outcomepoorelevated in liquid biopsies; associated with prognosisBustos, 2022120; Ordoñez-Rubiano, 202440miR-222serum (total/exosomal); CSFhigh → poor OS/DFSpoorOncomiR; higher levels predict decreased survivalOlioso, 2021102; Zhao, 2017116miR-328-3pserum EV (CD44+)high → poor outcomepoorincluded among increased EV-miRNAs linked to poor survivalTzaridis, 202080miR-485-3pserumlow → poor OSpoorlow serum level predicts shorter survivalWang, 2017108miR-487b-3pserumlow → poor OS/PFSpoorwith miR-1-3p & miR-26a-1-3p in serum signature; lower = worseDíaz Méndez, 202378miR-663serum/tissuelow → poor outcomepoorlower miR-663 associated with unfavorable clinical outcomeChen, 2017b123miR-2276-5pplasma (exosomes)low → poor OSpoortargets RAB13; low in exosomes associates with higher grade and shorter survivalSun, 2021112Associations with OS, PFS, and DFS are reported. CD44, cluster of differentiation 44; CNS, central nervous system; CSF, cerebrospinal fluid; DFS, disease-free survival; EV, extracellular vesicle; GBM, glioblastoma; IDHwt, isocitrate dehydrogenase wild-type; MGMT, O6-methylguanine-DNA methyltransferase; miR, microRNA; OS, overall survival; PFS, progression-free survival; RT, radiotherapy; TMZ, temozolomide. Circulating miRNAs with prognostic value in glioma Associations with OS, PFS, and DFS are reported. CD44, cluster of differentiation 44; CNS, central nervous system; CSF, cerebrospinal fluid; DFS, disease-free survival; EV, extracellular vesicle; GBM, glioblastoma; IDHwt, isocitrate dehydrogenase wild-type; MGMT, O6-methylguanine-DNA methyltransferase; miR, microRNA; OS, overall survival; PFS, progression-free survival; RT, radiotherapy; TMZ, temozolomide. Among upregulated oncomiRs, miR-21 correlates with shorter OS and higher recurrence risk across serum, plasma, and CSF exosomes.95,102,113,117 Other adverse prognostic markers include miR-222 (predicting reduced OS and disease-free survival [DFS]),102,116 miR-210 (reflecting tumor hypoxia),101 miR-182,116 miR-196b,41 miR-193b,122 miR-20a-5p, and miR-106a-5p.116 Exosomal signatures have revealed that high expression of miR-15b-3p, miR-21-3p, and miR-328-3p in CD44+ vesicles strongly associates with poor outcome,113 and miR-221 and miR-155 are adverse prognostic biomarkers across multiple biofluids.40,120 Conversely, reduced levels of tumor-suppressor miRNAs correlate with poor prognosis. Decreased miR-181a, miR-181b, and miR-181d have been associated with shorter OS, with miR-181d also predicting MGMT methylation status and treatment response.81,96,97 Other examples include miR-137,94 miR-100,92 miR-205,124 miR-185,121 miR-2276-5p,112 miR-485-3p,108 miR-29b,88 miR-203, miR-663,123 and miR-151a.119 Interestingly, low circulating miR-30e has paradoxically been associated with better OS.40 Composite signatures again outperform single miRNAs. A three-miRNA serum panel (miR-1-3p, miR-26a-1-3p, miR-487b-3p) stratified isocitrate dehydrogenase (IDH)-wild-type GBM patients and predicted survival,78 while vesicle-derived panels (miR-15b-3p, miR-21-3p, miR-328-3p) demonstrated superior prognostic discrimination.113 Integrated miRNA signatures hold the greatest promise for robust survival prediction. Monitoring treatment response Circulating miRNAs have been explored as dynamic biomarkers to monitor treatment response and disease progression. Several studies show that longitudinal changes in miRNA expression mirror therapeutic efficacy or herald recurrence (Table 5).Table 5Circulating miRNAs for monitoring treatment response in gliomamiRFluid/matrixExpression change vs. treatment courseTreatment typeClinical noteAuthors (name, year)miR-10bCSF/serum↑ associated with progression; decrease post-responseRT + TMZ; general follow-upstemness/invasion miRNA; higher levels track active diseaseKopková, 201941miR-21serum/plasma/CSF exosomes↓ in responders; ↑/remain high in non-responders; ↑ at progressionStupp protocol (RT + TMZ); Anti-angiogenic therapy (bevacizumab); general longitudinal follow-uptracks disease activity; dynamic across therapy; best-validated liquid markerHartanto, 2024125; Siegal, 2016126; Olioso, 2021102; Wang, 201283miR-124-3pserum exosomes↑ in progressive disease; normalization suggests responsegeneral longitudinal follow-uptumor suppressor in tissue; exosomal release increases with progressionOlioso, 2021102miR-128plasma/serumcandidate: ↓ with progression; ↑ with disease control (cohort-dependent)RT + TMZ; general follow-upneuronal identity; tissue suppressor—circulating dynamics reported in your draftWang, 201283miR-155plasma/serumdynamic across treatment; ↑ may herald progression; ↓ with responseRT + TMZ; general longitudinal follow-upinflammation/immune-related miRNA; included in longitudinal panelsWu, 202297miR-181dplasma/serum; Exosomeshigher baseline → better response; low/declining levels → poor responseRT + TMZ (MGMT-linked)predictive of MGMT status and treatment response; useful for monitoringSippl, 202295; Stakaitis, 202096miR-185plasmareported ↓ after therapy in responders (normalize with disease control)surgery + adjuvant therapy; RT + TMZcandidate longitudinal marker; needs standardized matricesTang, 2015121miR-210serum/Exosomes↓ with response (reduced hypoxia); ↑ with resistance/progressionRT + TMZ; bevacizumab (anti-angiogenic)hypoxia-related oncomiR; reflects tumor oxygenatioN/AngiogenesisLan, 2020101miR-222serum/CSF; EVs↓ in responders; ↑ in non-responders; rises with progressionStupp protocol (RT + TMZ); anti-angiogenic therapy (bevacizumab)parallels miR-21 dynamics; linked to DNA repair and drug resistanceHartanto, 2024125; Siegal, 2016; Olioso, 2021102miR-342-3pplasmacandidate: changes reported with treatment courseRT + TMZ; general follow-upsignal noted alongside miR-128 in responders vs. non-respondersWang, 201283miR-410plasmadynamic across treatment; trends with response (panel data)RT + TMZincluded with miR-155/181 family in longitudinal assessmentsWu, 202297 (panel)miR-485-3pserumreturns toward normal after surgery/therapy in responderssurgery + adjuvant therapylow levels associate with poorer outcome; dynamic normalization post-treatmentWang, 2017108miR-1238serum exosomes↑ in non-responders/TMZ-resistant; decreases with restored sensitivityTMZexosomal transfer mediates chemoresistance; candidate for real-time resistance trackingYin, 2019127Dynamic changes in miRNA expression are reported across surgical, chemoradiation, and anti-angiogenic therapy contexts. CSF, cerebrospinal fluid; EV, extracellular vesicle; MGMT, O6-methylguanine-DNA methyltransferase; miR, microRNA; RT, radiotherapy; TMZ, temozolomide. Table 5Circulating miRNAs for monitoring treatment response in gliomamiRFluid/matrixExpression change vs. treatment courseTreatment typeClinical noteAuthors (name, year)miR-10bCSF/serum↑ associated with progression; decrease post-responseRT + TMZ; general follow-upstemness/invasion miRNA; higher levels track active diseaseKopková, 201941miR-21serum/plasma/CSF exosomes↓ in responders; ↑/remain high in non-responders; ↑ at progressionStupp protocol (RT + TMZ); Anti-angiogenic therapy (bevacizumab); general longitudinal follow-uptracks disease activity; dynamic across therapy; best-validated liquid markerHartanto, 2024125; Siegal, 2016126; Olioso, 2021102; Wang, 201283miR-124-3pserum exosomes↑ in progressive disease; normalization suggests responsegeneral longitudinal follow-uptumor suppressor in tissue; exosomal release increases with progressionOlioso, 2021102miR-128plasma/serumcandidate: ↓ with progression; ↑ with disease control (cohort-dependent)RT + TMZ; general follow-upneuronal identity; tissue suppressor—circulating dynamics reported in your draftWang, 201283miR-155plasma/serumdynamic across treatment; ↑ may herald progression; ↓ with responseRT + TMZ; general longitudinal follow-upinflammation/immune-related miRNA; included in longitudinal panelsWu, 202297miR-181dplasma/serum; Exosomeshigher baseline → better response; low/declining levels → poor responseRT + TMZ (MGMT-linked)predictive of MGMT status and treatment response; useful for monitoringSippl, 202295; Stakaitis, 202096miR-185plasmareported ↓ after therapy in responders (normalize with disease control)surgery + adjuvant therapy; RT + TMZcandidate longitudinal marker; needs standardized matricesTang, 2015121miR-210serum/Exosomes↓ with response (reduced hypoxia); ↑ with resistance/progressionRT + TMZ; bevacizumab (anti-angiogenic)hypoxia-related oncomiR; reflects tumor oxygenatioN/AngiogenesisLan, 2020101miR-222serum/CSF; EVs↓ in responders; ↑ in non-responders; rises with progressionStupp protocol (RT + TMZ); anti-angiogenic therapy (bevacizumab)parallels miR-21 dynamics; linked to DNA repair and drug resistanceHartanto, 2024125; Siegal, 2016; Olioso, 2021102miR-342-3pplasmacandidate: changes reported with treatment courseRT + TMZ; general follow-upsignal noted alongside miR-128 in responders vs. non-respondersWang, 201283miR-410plasmadynamic across treatment; trends with response (panel data)RT + TMZincluded with miR-155/181 family in longitudinal assessmentsWu, 202297 (panel)miR-485-3pserumreturns toward normal after surgery/therapy in responderssurgery + adjuvant therapylow levels associate with poorer outcome; dynamic normalization post-treatmentWang, 2017108miR-1238serum exosomes↑ in non-responders/TMZ-resistant; decreases with restored sensitivityTMZexosomal transfer mediates chemoresistance; candidate for real-time resistance trackingYin, 2019127Dynamic changes in miRNA expression are reported across surgical, chemoradiation, and anti-angiogenic therapy contexts. CSF, cerebrospinal fluid; EV, extracellular vesicle; MGMT, O6-methylguanine-DNA methyltransferase; miR, microRNA; RT, radiotherapy; TMZ, temozolomide. Circulating miRNAs for monitoring treatment response in glioma Dynamic changes in miRNA expression are reported across surgical, chemoradiation, and anti-angiogenic therapy contexts. CSF, cerebrospinal fluid; EV, extracellular vesicle; MGMT, O6-methylguanine-DNA methyltransferase; miR, microRNA; RT, radiotherapy; TMZ, temozolomide. miR-21 levels typically decrease in responders following surgery, chemoradiation, or anti-angiogenic therapy but remain elevated or increase at progression.83,102,125,126 miR-222 demonstrates a similar pattern,102,125,126 and hypoxia-associated miR-210 decreases with effective therapy and re-elevates upon resistance, making it particularly relevant for bevacizumab monitoring.101,126 miR-10b in CSF and serum marks active disease,41 exosomal miR-124-3p rises in progressive disease,102 and miR-1238 mediates temozolomide resistance—being enriched in exosomes from resistant patients and decreasing when sensitivity is restored.127 Several tumor-suppressor miRNAs correlate with treatment benefit: higher miR-181d predicts better response to chemoradiation, consistent with its association with MGMT methylation81,96; normalization of miR-485-3p after therapy associates with favorable response108; and changes in miR-128 and miR-342-3p have been noted across treatment courses.83 Panel-based approaches improve discrimination between responders and non-responders. These results underscore the potential of circulating miRNAs as real-time indicators of treatment response. However, reproducibility across cohorts remains limited, and composite panels measured longitudinally are likely to offer the most reliable monitoring tools. Differentiating tumor recurrence from radionecrosis Post-operative monitoring remains challenging, as MRI can be confounded by pseudo-progression or pseudo-response, particularly after chemoradiation or anti-angiogenic therapy. The pivotal study by Morokoff et al. is, to date, the only systematic prospective investigation of serum miRNAs for post-operative glioma follow-up. In a cohort of 91 glioma patients (44 GBM, 47 low-grade glioma [LGG]) and 17 healthy controls, they identified a nine-miRNA serum signature with near-perfect accuracy for distinguishing glioma from healthy subjects. In longitudinal analysis of 11 patients with serial samples, miR-223 (in LGG) and miR-320e (in GBM) demonstrated dynamic fluctuations closely correlated with MRI-measured tumor volume. Critically, serum miRNA levels did not increase in cases of pseudo-progression—unlike MRI changes—suggesting that circulating miRNAs may provide additional specificity to distinguish true progression from treatment-related effects.128 Combining miR-223, miR-320e, miR-21, and miR-23a achieved diagnostic accuracy approaching 100%, with miR-223 and miR-320e emerging as the most promising markers for longitudinal follow-up (Table 6).Table 6Circulating miRNAs for differentiating true tumor recurrence from pseudo-progression128miRFluid/matrixExpression dynamicsTumor type/contextClinical noteAuthors (name, year)miR-223serum↓ after resection; ↑ at recurrenceLGGdynamic correlation with tumor volume; did not increase in pseudo-progressionMorokoff, 2020128miR-320eserum↓ after resection; ↑ at recurrenceGBMcorrelates with tumor volume; stable in pseudo-progressionMorokoff, 2020128miR-21serumcontributes to diagnostic 4-miR panel; dynamic levels reportedGBM/LGGcombined with miR-223, miR-320e, and miR-23a improves diagnostic accuracyMorokoff, 2020128miR-23aserumpart of diagnostic 4-miR panelGBM/LGGless dynamic individually, but increases accuracy when combinedMorokoff, 2020128Serum miRNA dynamics were assessed longitudinally in glioma patients with serial sampling. GBM, glioblastoma; LGG, low-grade glioma; miR, microRNA. Table 6Circulating miRNAs for differentiating true tumor recurrence from pseudo-progression128miRFluid/matrixExpression dynamicsTumor type/contextClinical noteAuthors (name, year)miR-223serum↓ after resection; ↑ at recurrenceLGGdynamic correlation with tumor volume; did not increase in pseudo-progressionMorokoff, 2020128miR-320eserum↓ after resection; ↑ at recurrenceGBMcorrelates with tumor volume; stable in pseudo-progressionMorokoff, 2020128miR-21serumcontributes to diagnostic 4-miR panel; dynamic levels reportedGBM/LGGcombined with miR-223, miR-320e, and miR-23a improves diagnostic accuracyMorokoff, 2020128miR-23aserumpart of diagnostic 4-miR panelGBM/LGGless dynamic individually, but increases accuracy when combinedMorokoff, 2020128Serum miRNA dynamics were assessed longitudinally in glioma patients with serial sampling. GBM, glioblastoma; LGG, low-grade glioma; miR, microRNA. Circulating miRNAs for differentiating true tumor recurrence from pseudo-progression128 Serum miRNA dynamics were assessed longitudinally in glioma patients with serial sampling. GBM, glioblastoma; LGG, low-grade glioma; miR, microRNA. Although limited by its relatively small longitudinal subset, this study provides proof of concept that serial monitoring of selected circulating miRNAs can complement MRI for post-operative surveillance. These findings strongly argue for larger multicentric validation studies. Integration with established molecular markers Contemporary glioma diagnosis and management rest on a molecular framework that has been substantially refined by the 2021 WHO Classification of Tumors of the Central Nervous System, in which IDH mutation status, 1p/19q codeletion, MGMT promoter methylation, and additional molecular features now define disease subtypes and inform therapeutic decisions.129 Any liquid biopsy candidate biomarker must therefore be evaluated against this tissue-based reference framework to determine whether it provides complementary or redundant information. IDH mutation status Although IDH mutations are detectable in plasma cell-free DNA in a subset of patients, this assay shows limited sensitivity in GBM. Several circulating miRNA signatures have been reported to differ according to IDH status; in particular, a three-miRNA serum panel (miR-1-3p, miR-26a-1-3p, miR-487b-3p) was specifically deregulated in IDH-mutant compared with IDH-wild-type gliomas and retained independent prognostic value.78 Such signatures may complement direct IDH genotyping by capturing downstream biological effects of IDH mutation, including the metabolic reprogramming triggered by 2-hydroxyglutarate accumulation. MGMT promoter methylation MGMT promoter methylation predicts temozolomide benefit but is assayed on tumor tissue, which is not always available, particularly at recurrence. miR-181d has emerged as the most consistently reported circulating biomarker correlated with MGMT methylation status and chemotherapy benefit,81,96 and the broader landscape of miRNAs regulating MGMT expression and modulating temozolomide sensitivity has been systematically reviewed.130 Circulating miR-181d levels could therefore serve as a surrogate marker when tumor tissue is unavailable. Molecular subtypes Beyond IDH and MGMT, GBM displays transcriptomic and methylation-based subtypes (proneural, classical, mesenchymal) with distinct biological behaviors and therapeutic vulnerabilities. Several miRNAs are differentially expressed across these subtypes—for instance, miR-21 is enriched in mesenchymal tumors, whereas miR-128 is downregulated in proneural tumors—suggesting that subtype-specific circulating signatures may eventually be defined to refine non-invasive glioma classification.131 Independent versus complementary value Whether circulating miRNAs provide truly independent prognostic information beyond IDH status, MGMT methylation, and clinical variables remains an open question. Both the Morokoff serum signature128 and the IDH-stratified three-miRNA panel78 retained predictive value in multivariable analyses, suggesting that circulating miRNAs contribute complementary rather than purely redundant information. Nevertheless, head-to-head comparisons with established molecular markers in large prospective cohorts are still lacking and represent a priority for future translational efforts. To enable a semi-quantitative comparison across studies, Table S1 ranks the circulating miRNAs most consistently validated in glioma in the present review, stratified by the number of independent supporting studies. miR-21 emerges as the most reproducible candidate, followed by miR-222, miR-210, the miR-181 family, miR-182, and miR-10b. By contrast, several promising candidates, including miR-223, miR-320e, and the exosomal miR-1238, currently rest on single high-quality studies and warrant independent validation before clinical adoption. Synthesis Figure 1 provides an integrative overview of circulating miRNAs detected in serum/plasma, CSF, and exosomes, stratified by clinical application. By mapping upregulated oncomiRs and downregulated tumor suppressors to diagnostic, prognostic, and monitoring contexts, it illustrates how liquid biopsy approaches may complement conventional imaging and histopathological evaluation.Figure 1Integrative overview of circulating miRNAs detected in serum/plasma, CSF, and exosomes, stratified by clinical application (diagnosis, prognosis, and monitoring)OncomiRs are shown in red (upregulated); tumor-suppressor miRNAs are shown in blue (downregulated). Integrative overview of circulating miRNAs detected in serum/plasma, CSF, and exosomes, stratified by clinical application (diagnosis, prognosis, and monitoring) OncomiRs are shown in red (upregulated); tumor-suppressor miRNAs are shown in blue (downregulated). Discussion Limitations Despite the promising utility demonstrated by circulating miRNAs in GBM monitoring, several critical limitations temper their clinical integration. First, significant pre-analytical and technical inconsistencies exist across studies. Variation in sample types (serum versus plasma versus CSF), collection methods, and RNA extraction protocols leads to inconsistent results, and normalization strategies lack consensus, as endogenous controls may themselves be altered in disease states.132 Second, the biological complexity and origin ambiguity of circulating miRNAs represent a major challenge. These molecules can originate not only from tumor cells but also from reactive stromal or immune cells, especially via EVs. Although EVs offer a protected route for miRNA dissemination, their isolation may co-purify contaminants, and their cargo heterogeneity reflects diverse cellular processes rather than tumor biology alone.133 This source-specificity issue is particularly acute in GBM, where circulating miRNAs reflect contributions from at least four distinct compartments: tumor cells; the tumor microenvironment (reactive astrocytes, microglia/macrophages, endothelial cells); peripheral immune cells, whose miRNA repertoires shift with systemic inflammation and treatment134; and platelets, which release abundant miRNAs upon activation and during serum coagulation. Because mature blood cells exhibit highly abundant and lineage-specific miRNA profiles, even minor hemolysis or incomplete platelet depletion during sample processing can mask or distort tumor-derived signals, and quality-control panels have been proposed to flag such contamination in clinical samples.135 This is particularly relevant for candidate biomarkers such as miR-223, which is enriched in granulocytes and platelets and may fluctuate with treatment-associated changes in blood cell counts independently of tumor biology. The BBB adds a further layer of complexity: in GBM, BBB disruption is heterogeneous in time and space and is likely to modulate the passage of tumor-derived miRNAs into the systemic circulation, contributing to inter-study variability and reinforcing the appeal of CSF as a tumor-proximal compartment less affected by these confounders. Emerging strategies to mitigate these limitations include surface-marker-based enrichment of tumor-derived EVs and focused ultrasound-induced BBB opening to facilitate the release of brain tumor biomarkers into the bloodstream.136 Third, many studies to date are based on small, single-institution cohorts, which may not reflect broader patient diversity or capture longitudinal dynamics. The absence of large-scale prospective validation significantly diminishes the generalizability of identified miRNA signatures.133 Fourth, analytical platform constraints further complicate clinical translation. As noted by Godoy et al., RT-qPCR arrays show high sensitivity for known targets and good reproducibility, but microarrays result in higher false-positive rates and correlate poorly with qPCR, while RNA-seq remains costly and suffers from standardization challenges.137 Additionally, electrochemical miRNA biosensors provide rapid and target-specific detection at lower cost, yet their clinical translation remains limited by issues of reproducibility, signal specificity, and cost-effectiveness.138 Finally, no circulating miRNA has been validated head to head against MRI or established molecular markers such as MGMT methylation in randomized clinical trials. Regulatory acceptance will require evidence of improved clinical decision-making through prospective, multicenter studies.133 Future perspectives Several avenues hold promise for advancing circulating miRNAs toward clinical implementation. On the technological front, novel multiplexed biosensors leveraging CRISPR-Cas systems, nanostructured probes, and microfluidic platforms are emerging as highly sensitive, rapid, and low-cost approaches for circulating miRNA detection, with potential for point-of-care GBM monitoring applications.139,140 The integration of miRNA profiles with other liquid biopsy analytes (e.g., ctDNA), proteomic markers, imaging features, and clinical variables—analyzed via machine-learning algorithms—could yield robust, dynamic biomarker signatures that better stratify patients and guide therapeutic decisions.141 In parallel, exosome cargo holds dual potential: as stable carriers of tumor-specific miRNAs for diagnostic applications, and as natural vehicles for delivering therapeutic oligonucleotides across the BBB, thereby potentially overcoming a major delivery challenge in GBM treatment.133,142 To bridge bench-to-bedside translation, standardized protocols for sample handling, miRNA quantification, and data analysis must be established. Large-scale, multicenter trial designs incorporating miRNA endpoints alongside radiographic and molecular metrics will be essential to confirm clinical utility and achieve regulatory acceptance. In summary, while circulating miRNAs offer a compelling path toward non-invasive, dynamic GBM monitoring, technical, biological, and clinical hurdles must be addressed. Advances in detection technology, multi-omic integration, and rigorous validation frameworks represent the most promising avenues to realize their full translational potential—moving toward personalized, real-time monitoring in neuro-oncology. Conclusion Circulating miRNAs represent highly promising biomarkers for glioma management, with evidence supporting their role across diagnosis, prognosis, treatment monitoring, and recurrence assessment. However, significant challenges remain, particularly regarding technical standardization and prospective validation. Integration with multi-omic platforms, advanced detection technologies, and clinical-trial frameworks may ultimately pave the way toward their adoption in precision neuro-oncology. Materials and methods Search strategy This review comprises two parts: a narrative overview of circulating miRNA biology, pre-analytical handling, and detection methods; and a systematic review conducted according to PRISMA guidelines to identify clinical studies on circulating miRNAs as biomarkers in glioma. Literature searches were performed in PubMed, Embase, Scopus, and Web of Science using terms combining “glioma” or “glioblastoma” with “microRNA,” “miRNA,” and “circulating.” Boolean operators and truncation were applied. The search was limited to peer-reviewed articles published in English from January 2000 to August 2025. Additional records were identified from reference lists of selected studies. All references were imported into a citation manager and duplicates removed. Inclusion and exclusion criteria Studies were included if they (1) reported original data on circulating miRNAs in glioma patients; (2) used blood (serum, plasma) or CSF as the miRNA source; (3) reported diagnostic, prognostic, predictive, or monitoring outcomes; and (4) were published in peer-reviewed English-language journals. Studies based exclusively on tissue miRNAs, non-original works (reviews, editorials, case reports, conference abstracts), studies without a clinical cohort, and those with insufficient methodological detail were excluded. Study selection and data extraction Titles and abstracts were independently reviewed by two investigators. Full texts of potentially relevant articles were retrieved for detailed evaluation; disagreements were resolved by consensus or consultation with a third reviewer. Data were extracted using a standardized form covering study characteristics (first author, year, country, design, sample size), patient characteristics (age, sex, histological and molecular subtype, treatment context), biofluid source, miRNA(s) analyzed, detection method, normalization strategy, and clinical outcomes. Due to heterogeneity in study design, patient cohorts, biofluid sources, and outcome measures, a quantitative meta-analysis was not feasible. Findings were therefore synthesized narratively, emphasizing recurring miRNA candidates, study quality, and consistency of results. The initial PubMed search yielded 512 records. After duplicate removal, 489 records were screened, of which 409 were excluded. Eighty full-text articles were assessed; 16 were excluded (six without circulating miRNA analysis, four including non-glioma populations, five lacking clinical outcomes, and one duplicate dataset). Sixty-four studies met the inclusion criteria (Figure 2).Figure 2PRISMA flow diagram of the systematic literature search and study selection process. PRISMA flow diagram of the systematic literature search and study selection process. Study quality assessment Given the marked heterogeneity of available studies and the absence of a single validated risk-of-bias tool tailored to circulating miRNA biomarker research, a pragmatic qualitative grading framework was applied to characterize the methodological strength of each included study rather than to formally pool or exclude studies. Five complementary dimensions were considered: (1) cohort size, stratified as small (<50 patients), moderate (50–100), or large (>100); (2) study design, distinguishing prospective from retrospective enrollment; (3) inclusion of an independent validation cohort or external dataset; (4) control of pre-analytical and technical variables, including hemolysis assessment, normalization strategy, and standardized RNA extraction protocols; and (5) statistical rigor, with multivariable analyses controlling for known prognostic factors (age, KPS, IDH/MGMT status, treatment) considered superior to univariate analyses alone. These criteria informed the qualitative weighting of evidence throughout the present review and underlie the consistency tiers reported in Table S1. Particular emphasis was placed on candidate miRNAs whose performance was replicated across independent cohorts, biofluid types, and detection platforms. Conversely, results derived exclusively from small, single-institution, retrospective cohorts without independent validation are flagged in the narrative as preliminary and requiring confirmation in larger prospective studies. This framework should be regarded as an informal and pragmatic appraisal rather than as a formal risk-of-bias assessment such as ROBINS-I or QUADAS-2, which were not directly applicable given the predominantly observational and biomarker-discovery nature of the included studies. Data and code availability This study did not generate new data or code. All data discussed in this review were obtained from previously published studies, which are cited in the reference list and are publicly available through their original publications.