The appeal of these compartments lies in basic physics and biology. Plasma-based liquid biopsy has transformed oncology by detecting circulating tumor DNA, or ctDNA, the fragments of tumor genome released into the bloodstream when cancer cells die through apoptosis or necrosis. But blood is a highly diluted medium, contaminated with abundant DNA from dying white blood cells, and the blood-brain barrier actively excludes tumor DNA shed by metastases in the brain or its protective membranes. Fluids that sit in direct contact with tumor deposits face neither problem. Ascites bathing peritoneal metastases, pleural fluid in contact with lung tumors, and cerebrospinal fluid surrounding the central nervous system all receive tumor-derived DNA at high local concentrations, with far less background genetic noise from peripheral blood cells. The result is a signal that is not only stronger but, in some cases, fundamentally different from what blood or a single needle biopsy of tissue can provide.

The research team, led by investigators including Aditya Shreenivas and Razelle Kurzrock, profiled fluid samples using a next-generation sequencing platform capable of detecting point mutations, copy number alterations, splice variants, indels, and gene fusions. The cohort of 45 patients, with a median age of 52, included 18 men and 27 women, and spanned lung cancer, breast cancer, ovarian cancer, pancreatic cancer, and a collection of rarer malignancies. Ascitic fluid accounted for 26 samples, pleural fluid 11, cerebrospinal fluid seven, and pericardial fluid one. The overall ctDNA detection rate was a striking 87 percent, with ascites performing best at 96 percent, pleural fluid at 72.7 percent, and cerebrospinal fluid at 71.4 percent. The single pericardial sample was also positive, though the authors caution that one case proves little. These rates, while descriptive rather than definitive performance estimates, underline how readily tumors seed their genetic material into the cavities they invade.

The mutational landscape the sequencing revealed was dominated by familiar oncogenic players. TP53, the guardian-of-the-genome gene whose loss is a hallmark of nearly all cancers, was altered in 49 percent of samples. EGFR and KRAS, two of the most clinically consequential oncogenes in lung and gastrointestinal malignancies, were each altered in 18 percent of patients, with co-existing EGFR amplifications detected in three cases that also carried sensitizing EGFR mutations. PIK3CA and FGFR3 each appeared in 9 percent of patients, GNAS and MYC in 7 percent, and CTNNB1 and the estrogen receptor gene ESR1 in 4 percent each. Amplifications were present in 20 percent of the cohort, most frequently affecting EGFR and MYC. No gene fusions or rearrangements were detected, a negative finding the authors report explicitly.

Perhaps the most clinically electrifying numbers came from the variant allele frequency analysis, a measure of what fraction of the cell-free DNA in a sample carries the mutation in question. Cerebrospinal fluid posted a median VAF of 23.2 percent, an order of magnitude higher than the 0.65 percent median typically achievable in plasma, reflecting the fluid’s remarkable purity and minimal contamination from blood-derived genomic DNA. Pleural fluid showed a median VAF of 6.5 percent and ascites 1.23 percent, while the lone pericardial sample harbored an EGFR mutation at a VAF of 70 percent. Such high fractions mean that even modest sequencing depth can reliably detect clinically relevant alterations in these fluids, potentially enabling the identification of low-frequency mutations and structural variants that current plasma-based platforms struggle to resolve.

To translate raw mutation lists into clinical meaning, the researchers classified every alteration using the ESCAT framework, the European Society for Medical Oncology’s scale for ranking the actionability of molecular targets. Roughly 31 percent of patients carried tier IA alterations, meaning mutations with approved targeted therapies for their specific tumor type, almost all of them in breast or lung cancer. Another 35.6 percent fell into tier IIIA, where drugs have proven benefit in the same molecular subtype but a different tumor type, and 53.3 percent carried tier IVA alterations supported only by preclinical evidence, highlighting unmet opportunities for clinical trial enrollment. The take-home message is that fluid profiling is not merely academic; the alterations it finds map directly onto therapies that already exist or are in development.

Even more provocative were the six patients whose fluids harbored therapeutically relevant alterations that were entirely absent from matched tissue and blood samples. One woman with breast cancer had an ERBB2 amplification detectable only in her ascites, a finding that opens the door to multiple FDA-approved HER2-directed drugs. Another breast cancer patient carried the ESR1 D538G resistance mutation, targetable with agents including elacestrant and fulvestrant, in her cerebrospinal fluid alone. A gastrointestinal stromal tumor patient had a KIT mutation in pleural fluid that imatinib can inhibit, and three patients, one with ovarian and two with pancreatic cancer, showed FGFR3 amplifications or KRAS G12V mutations exclusively in ascitic fluid. These discordances are the molecular fingerprint of spatial and temporal tumor heterogeneity: different metastatic sites evolve different resistance mechanisms, and local shedding enriches the adjacent fluid with clones that a single tissue biopsy or diluted plasma sample simply cannot see.

The concordance analysis quantified this incompleteness of conventional sampling. Among 16 patients with both fluid and blood profiling, only half shared at least one oncogenic alteration between the two sample types. Of 11 patients with matched tissue comparisons, seven showed overlapping alterations, and when all three sample types were compared simultaneously in seven patients, just three displayed fully concordant profiles. Cytology comparisons told a complementary story: of five cytology-negative fluid samples, four were positive for ctDNA, meaning molecular analysis flagged malignancy that the microscope had missed. The reverse also occurred, with three of twenty cytology-positive samples testing negative by sequencing, a pattern that reinforces the authors’ argument that the two approaches are complementary rather than competing. Fluid cytology retains formal staging importance, contributing, for example, to FIGO stage III classification in ovarian cancer, but ctDNA analysis catches cases cytology alone would miss.

The study is candid about its limitations. It was retrospective, exploratory, and descriptive, with a small and heterogeneous cohort, no formal statistical hypothesis testing, and representation of pericardial fluid limited to a single case. Matched samples were not available for every patient, timing differences between fluid, tissue, and plasma collection may have contributed to discordance through clonal evolution under therapeutic pressure, and platform differences in panel design and sequencing depth cannot be fully excluded as sources of the observed discrepancies. ESCAT tiers link drugs to mutations but do not demonstrate that any individual patient benefited from matching therapy. The authors explicitly frame their detection rates as preliminary observations that require prospective validation in larger, disease-specific cohorts with standardized sampling.

Even with those caveats, the implications are difficult to ignore. Effusions and cerebrospinal fluid are collected routinely from advanced cancer patients, often repeatedly, yet their genomic content has been largely discarded after the cytology slide is read. The new findings, consistent with prior reports of EGFR amplification found in ascites but not tissue, and with NCCN guidelines that already recommend considering CSF tumor DNA testing when leptomeningeal disease is suspected, suggest that these fluids deserve a formal place in precision oncology workflows. Body fluids sidestep the clonal hematopoiesis false positives that plague plasma sequencing, they can be sampled when tissue biopsy is unsafe or infeasible, and they capture the regional biology of metastatic sites that governs how each deposit will respond to treatment. If prospective trials confirm these descriptive signals, the cloudy fluid drained from a patient’s chest or abdomen may become as informative as any gene panel, transforming a routine byproduct of cancer care into a real-time map of the tumor’s evolving vulnerabilities.

Subject of Research: Circulating tumor DNA analysis of non-blood body fluids for cancer detection and molecular profiling

Article Title: Circulating tumor DNA in body fluids for molecular characterization of cancers beyond tissue and blood

Article References: Circulating tumor DNA in body fluids for molecular characterization of cancers beyond tissue and blood. (n.d.). https://doi.org/10.1016/j.isci.2026.117893

Image Credits: AI Generated

DOI: 10.1016/j.isci.2026.117893

Keywords: liquid biopsy, circulating tumor DNA, ascites, cerebrospinal fluid, pleural effusion, cytology, precision oncology, tumor heterogeneity, ESCAT, next-generation sequencing, EGFR, ERBB2