Urine and spinal fluid biomarkers show promise for detecting glioma early

Patients lose an average of 20 years of life, and the five-year survival rate across all gliomas is 35.7 percent. For glioblastoma, the most aggressive subtype, that figure falls to 6.9 percent. Diagnosis currently relies on magnetic resonance or computed tomography imaging followed by biopsy, and a protracted diagnostic pathway may delay effective care. A new systematic review now takes stock of a minimally invasive alternative: liquid biopsies drawn not from blood, but from two underexplored body fluids, urine and cerebrospinal fluid.

The researchers searched three electronic databases, MEDLINE, Embase and BIOSIS, from inception to September 2024, and screened 5209 records. After deduplication and full-text assessment of 506 articles, 20 studies met the inclusion criteria, covering a combined total of 669 glioma patients and 597 controls. Five studies examined urine-derived biomarkers and 15 examined biomarkers in cerebrospinal fluid, the clear fluid that bathes the brain and spinal cord.

The logic behind comparing these two fluids is biological. Urine can be collected non-invasively in large volumes and contains relatively little background protein, which may make low-abundance tumour signals easier to spot. Cerebrospinal fluid, by contrast, sits in direct contact with the brain tumour microenvironment and may therefore carry tumour-derived molecules at higher concentrations. Blood was excluded because it has already been comprehensively reviewed elsewhere. To be included, studies had to be cohort, case-control or cross-sectional in design, report non-blood biomarkers for glioma detection of any grade, include histologically confirmed glioma graded under the WHO system, and enrol at least 10 cases and 10 controls.

The urinary findings are striking on paper. Hallal and colleagues identified 31 differentially expressed proteins between 14 healthy controls and 24 patients with grade 4 glioblastoma; nine proteins individually showed an area under the ROC curve of at least 0.9, and a five-protein panel of KRT19, RPS2, RPL18, RPL28 and RPL7A achieved an AUC of 0.958. Wu and colleagues found a six-protein panel distinguishing glioma patients from healthy controls with an AUC of 0.96, sensitivity of 90 percent and specificity of 92 percent. Metabolomic studies added further candidates: Tandle and colleagues reported that a 46-metabolite model distinguished glioblastoma patients from controls with 74 percent sensitivity and 79 percent specificity, while Shi and colleagues identified urinary histidine with an AUC of 0.91. A study of urinary cell-free DNA by Mouliere and colleagues found that fragments between 30 and 60 base pairs were significantly more abundant in glioma patients and classified groups with an AUC of 0.89.

Yet the review’s authors urge heavy caution. Only five urinary studies met the criteria, all with fewer than 100 cases, and although four of the five quoted AUCs above 0.9, confidence intervals were not reported. The urinary protein and metabolite signals largely reflect disrupted amino acid, carbohydrate and nucleotide metabolism, inflammation and structural remodelling, which may represent a generalised response to disease rather than glioma-specific pathology. Histidine, for example, is a precursor of the inflammatory marker histamine and has been linked to breast and colorectal cancers, making it unlikely to be a specific cancer marker. The review found no overlap between reported urinary biomarkers and those reported in blood or cerebrospinal fluid, leaving unclear the extent to which urine signals detect specific glioma biology.

Cerebrospinal fluid yielded a richer and more varied literature. MicroRNAs, small non-coding RNAs that circulate as cancer biomarkers, were the most frequently reported biomarker class, appearing in six of twelve CSF studies. MiR-21 was the single most reported biomarker: four studies found significantly elevated CSF concentrations in glioma patients across WHO grades 2 to 4, and two reported accuracy statistics. Akers and colleagues reported sensitivities of 85 to 87 percent and specificities of 93 to 100 percent across two glioblastoma cohorts, while Qu and colleagues reported an AUC of 0.81. Baraniskin and colleagues tested a two-step classification tree using miR-15b followed by miR-21, reaching 90 percent sensitivity and 100 percent specificity, and Teplyuk and colleagues used a support vector machine informed by seven miRNAs to correctly classify 91.2 percent of samples as glioblastoma or control.

Protein and multi-omics classifiers in cerebrospinal fluid also performed well in individual studies. A proteomic classifier built by Mikolajewicz and colleagues distinguished glioblastoma from primary central nervous system lymphoma, brain metastases and hydrocephalus with an AUC of 0.95, sensitivity of 89 percent and specificity of 100 percent, although the individual protein GAP43 alone reached only an AUC of 0.69. Landry and colleagues combined proteome and methylome profiling to separate glioma from brain metastases and lymphoma, with three modelling approaches achieving AUCs between 0.85 and 0.89. Metabolomic profiling by Möhn and colleagues identified six metabolites, including putrescine and glutamate, at higher concentrations in glioma patients, each with individual AUCs between 0.728 and 0.785. Vascular endothelial growth factor and interleukin-8 were significantly elevated in glioma patients, while the myelin protein Nogo-A was, counterintuitively, ten-fold lower in glioma patients than in controls.

Formal meta-analysis proved impossible because the biomarkers, analysis methods and patient demographics differed too much across studies, so the findings were synthesised narratively. Risk of bias assessment with the QUADAS-2 tool revealed substantial methodological weaknesses: ten of nineteen assessable studies had a high risk of bias in patient selection, and all used a two-gate, case-control design recruiting patients with known disease and controls without it. Such designs are known to inflate estimates of diagnostic accuracy. Moreover, roughly 89 percent of participants across the studies had high-grade glioma, grades 3 or 4, meaning the evidence says little about genuinely early detection. More than half of the studies failed to fully report patient demographics, including sex and age, which may matter if biomarker abundance differs between sexes or across the lifespan.

Sampling itself poses a barrier to clinical translation. Cerebrospinal fluid is typically obtained by lumbar puncture or collected during neurosurgery, and concentrations of certain miRNAs and proteins differ significantly between lumbar and ventricular or cisternal samples. In a telling example from the review, the same nine-miRNA panel developed by Akers and colleagues achieved 80 percent sensitivity when tested on cisternal fluid from ten glioblastoma patients but only 28 percent sensitivity on lumbar fluid from eight patients, with specificities of 67 and 95 percent respectively. Lumbar puncture is also contraindicated for some patients with space-occupying lesions and raised intracranial pressure, and ventricular fluid is usually sampled opportunistically during surgery, limiting the practical role of CSF biomarkers to patients whose tumours cannot readily be biopsied.

The review’s conclusion is measured: urine and cerebrospinal fluid remain under-researched for glioma liquid biopsy, with existing studies limited by small, underpowered cohorts and inadequate case-control design. Several biomarkers, particularly miR-21, show promise replicated across more than one study, and multiple panels report AUCs above 0.9, but these estimates lack precision intervals and come from small samples. The authors call for well-designed, adequately powered studies with appropriate comparator groups before any of these biomarkers can enter a diagnostic pathway. For now, the message is one of cautious optimism: two accessible body fluids may eventually help catch glioma earlier, but the evidence base is still too thin to say how well, or for whom.

Subject of Research: Urine and cerebrospinal fluid biomarkers for early detection of glioma

Article Title: Urine and cerebrospinal fluid biomarkers for early glioma detection: a systematic review

Article References: Paterson, C., Yu, J., Davies, P., Read, J., Palmer, J. C., Dawson, S., Higgins, J. P., & Kurian, K. M. (2026). Urine and cerebrospinal fluid biomarkers for early glioma detection: a systematic review. Diagnostic and Prognostic Research, 10(1), Article 16. https://doi.org/10.1186/s41512-026-00230-1

Image Credits: AI Generated

DOI: 10.1186/s41512-026-00230-1

Keywords: glioma, liquid biopsy, cerebrospinal fluid, urine biomarkers, miR-21, cell-free DNA, glioblastoma, systematic review, diagnostic accuracy, QUADAS-2, metabolomics, proteomics