Colorectal cancer (CRC) ranks third in global cancer incidence, with roughly 1.93 million new cases and 935,000 deaths recorded in 2022. Prognosis depends heavily on the stage at which cancer is found: patients diagnosed at stage I have a five-year survival rate near 92%, while those diagnosed at stage IV face a rate below 14%.
Current screening tools fall short. Colonoscopy, the gold standard, is invasive and requires bowel preparation, leading to poor patient compliance. Stool-based tests such as the fecal immunochemical test (FIT) offer better convenience but detect fewer than 50% of stage I cancers. The multitarget stool DNA test Cologuard reaches 74% sensitivity for stage I, still leaving a substantial gap.
Liquid biopsy has emerged as a promising alternative: the analysis of blood or other body fluids for cancer-related molecules. It is non-invasive, can be repeated frequently, and can capture molecular signals from a tumor even before it becomes visible on imaging or endoscopy. The key targets in CRC liquid biopsy are circulating tumor DNA, circulating tumor cells, extracellular vesicles, and circulating proteins or microbial markers.
Circulating tumor DNA (ctDNA) consists of small DNA fragments shed from cancer cells into the bloodstream. In early CRC, ctDNA makes up only 0.01 to 0.1% of all cell-free DNA in blood, making it challenging but crucially informative to detect. Its short half-life of roughly 16 minutes to 2.5 hours means it reflects current tumor activity in near real time.
The most well-studied ctDNA markers are genetic mutations in driver genes. The APC gene is mutated in about 80% of CRC cases. One large study using ctDNA detection of APC, KRAS, and BRAF mutations achieved 87% sensitivity for stage I CRC, with a specificity of 99.3%. KRAS mutations (present in roughly 40% of cases) detected by digital droplet PCR caught 78% of stage I tumors, outperforming FIT.
Epigenetic changes, specifically DNA methylation, offer a more stable and earlier signal than mutations. The SEPT9 methylation test (Epi proColon) was the first FDA-approved ctDNA-based screening assay and achieves 73% sensitivity for stage I CRC. Combining SEPT9 with NDRG4 methylation detection boosts stage I sensitivity to 85% with 95% specificity, because methylation changes in tumor suppressor genes occur even before genetic mutations accumulate.
A key limitation of ctDNA testing is that results depend heavily on technical variables such as blood collection methods, processing time, and storage conditions. Most published studies are retrospective and involve fewer than 500 patients, meaning large-scale prospective trials are still needed before widespread clinical use.
Circulating tumor cells (CTCs) are intact cancer cells shed from a tumor and found traveling through the bloodstream. In early CRC, there are typically only 1 to 10 CTCs per milliliter of blood, compared to roughly one billion normal blood cells, making their isolation technically demanding. CTCs retain the molecular features of the original tumor, making them valuable for both early detection and treatment planning.
The FDA-approved CellSearch system captures CTCs using antibodies against the EpCAM protein on cancer cell surfaces. However, it detects only 30 to 40% of stage I CRC patients because some early cancer cells downregulate EpCAM as they begin to spread. Newer size-based methods like ISET catch 60 to 65% of stage I patients, and the microfluidic CTC-iChip reaches 65 to 70%, by combining physical and immunological separation techniques.
Beyond simply counting CTCs, molecular profiling of individual cells adds diagnostic power. Detecting KRAS mutations directly in CTCs achieves 72% sensitivity for stage I CRC, while identifying specific gene expression signatures such as overexpression of MACC1 can distinguish early CRC from benign bowel conditions with 80% sensitivity and 90% specificity.
The field faces standardization challenges: different platforms target different CTC subtypes, and CTCs degrade rapidly outside the body, requiring blood samples to be processed within 4 to 6 hours of collection. Until uniform protocols and head-to-head technology comparisons are established, clinical adoption remains limited.
Extracellular vesicles (EVs) are microscopic, membrane-bound packets released by cancer cells into the bloodstream. They carry a diverse cargo including proteins, small RNA molecules called microRNAs, and other genetic material, making them a rich source of early cancer signals. Their lipid membrane protects the cargo from degradation, improving measurement reliability.
EV-derived proteins such as CEA and EpCAM are enriched in vesicles shed by CRC cells and can be detected at earlier disease stages than free-circulating protein versions. The combined EV-CEA/EpCAM panel detects 82% of stage I CRC with 93% specificity, outperforming measurement of either protein alone in blood plasma.
EV-derived microRNAs are particularly promising. miR-21 is overexpressed in about 70% of early CRC and promotes cell growth, while miR-92a is overexpressed in roughly 80% of cases and supports tumor blood vessel formation. A large meta-analysis of 5,000 patients showed the combined EV-miR-21/92a panel achieves 80% sensitivity for stage I CRC with 92% specificity. Expanded RNA signatures including five microRNAs push stage I sensitivity to 85%.
A significant obstacle to clinical translation is the lack of standardized EV isolation methods. Ultracentrifugation, polymer precipitation, and size-exclusion chromatography each yield different EV preparations, making results hard to compare across laboratories. The International Society for Extracellular Vesicles has published guidelines, but clinical adoption remains inconsistent.
Beyond EVs, individual proteins circulating freely in blood have been studied as CRC markers. The classic CEA protein is widely used but has less than 50% sensitivity for stage I CRC. Newer markers show more promise: osteopontin (OPN), elevated in about 70% of early CRC, reaches 72% sensitivity for stage I, while matrix metalloproteinase-9 (MMP-9) achieves 68% sensitivity. Combining CEA with CA19-9 improves stage I detection to 65%.
A novel approach leverages the gut microbiome as a source of cancer signals. The bacterium Fusobacterium nucleatum colonizes CRC tissues and promotes tumor growth by activating Wnt signaling. Cell-free microbial DNA from F. nucleatum can be detected in blood and was found in 75% of stage I CRC patients with 90% specificity in a 2023 study.
Short-chain fatty acids (SCFAs) produced by healthy gut bacteria are reduced in early CRC due to microbiome disruption. Specifically, blood levels of butyrate, which has anti-inflammatory and anti-cancer properties, are 50% lower in stage I CRC patients than in healthy individuals. Combining F. nucleatum DNA detection with butyrate measurement achieves 82% sensitivity for stage I disease.
Microbial markers represent an emerging area with early-stage evidence, but gut microbiome composition varies substantially across populations due to diet, antibiotic use, and genetics. Larger validation studies across diverse groups will be needed before these markers can be recommended for routine screening.
Digital droplet PCR (ddPCR) works by dividing a blood sample into thousands of tiny droplets and testing each one for a specific DNA mutation. Because quantification is based on simple counting rather than comparison to a standard curve, ddPCR can detect mutations present at frequencies as low as 0.001%, making it ideal for picking up rare cancer signals. A 2024 study using ddPCR to find APC mutations in ctDNA detected 80% of stage I CRC with results available in under 24 hours. Its limitation is that each test targets only one or two specific known mutations.
Next-generation sequencing (NGS) reads millions of DNA fragments simultaneously, allowing detection of mutations, copy number changes, and methylation patterns across dozens of genes at once. Targeted NGS panels covering CRC-related genes can detect 85% of stage I CRC with 99% specificity. While comprehensive, NGS costs $500 to $1,000 per sample and takes 3 to 7 days, limiting its use in routine screening settings.
Bisulfite sequencing detects methylation changes in tumor suppressor genes by chemically converting unmethylated DNA while leaving methylated sites intact. A 2024 study using reduced representation bisulfite sequencing (RRBS) to analyze SEPT9 and NDRG4 methylation detected 83% of stage I CRC with 94% specificity. The main challenge is that bisulfite treatment degrades roughly 90% of the starting DNA, requiring larger input amounts and careful sample preparation.
Each technology has trade-offs: ddPCR is fast and inexpensive but narrow in scope; NGS is comprehensive but costly and slow; bisulfite sequencing excels at detecting methylation but requires high-quality samples. All three struggle with the extremely low concentrations of ctDNA present in stage 0 to I disease, where the cancer signal can fall at or below detection limits.
No single biomarker achieves the sensitivity and specificity needed for population-level CRC screening, which has driven research toward multi-marker panels that combine ctDNA methylation, EV-derived microRNAs, and circulating proteins. Preliminary studies show that integrating SEPT9/NDRG4 methylation with EV-miR-21/92a and EV-associated CEA can detect over 90% of stage I CRC, because each marker class captures cancer signals that the others miss.
Artificial intelligence (AI) and machine learning (ML) are essential for extracting meaningful patterns from these complex multi-analyte datasets. A 2024 study developed a deep learning model combining ctDNA methylation and EV-miRNA data, achieving 92% sensitivity for stage I CRC, 96% sensitivity for advanced precancerous polyps, and 98% specificity. Graph neural networks modeling interactions between different biomarkers achieved 91% sensitivity, while convolutional neural networks analyzing ctDNA fragment size patterns achieved 89%.
AI tools are being applied across the entire liquid biopsy workflow: discovering new biomarkers from large datasets using unsupervised learning, filtering technical noise to detect rare cancer signals, integrating multi-omics data into unified risk predictions, and forecasting which patients will respond to specific treatments. Transformer models adapted from language processing have been applied to ctDNA mutation signatures with 87% sensitivity for stage I CRC.
Significant challenges remain before AI-driven liquid biopsy can be used clinically. Models require large, diverse training datasets but most liquid biopsy studies involve small, single-center cohorts. Deep learning models can be difficult to interpret, raising concerns for regulatory approval. Models trained on one population may perform poorly in another due to genetic and environmental differences, underscoring the need for external validation in diverse groups.
Point-of-care (POC) devices aim to make liquid biopsy accessible outside major medical centers. Miniaturized microfluidic chips can detect KRAS mutations in ctDNA within one hour at a cost of only $50 per sample, compared to $200 to $500 for laboratory-based methods. Lateral flow assays similar to pregnancy tests can detect protein markers such as CEA or CA19-9 within 15 to 30 minutes, though with lower sensitivity than laboratory platforms (typically 60 to 70% for early-stage CRC).
Smartphone-based detection systems that read colorimetric or fluorescent signals from microfluidic chips may enable quantitative analysis without expensive equipment, and paper-based devices could reduce per-test costs to less than $1 in resource-limited settings. These innovations could help bring liquid biopsy screening to populations currently unable to access regular colonoscopy.
Multiple clinical scenarios are being evaluated for liquid biopsy integration: as a first-line screening test for average-risk adults aged 45 to 75, as a tool to identify who most urgently needs colonoscopy, as surveillance for patients who have had precancerous polyps removed, or as a triage test to reduce unnecessary colonoscopies after positive FIT results. Each scenario has different requirements for sensitivity and cost-effectiveness.
Health economic modeling suggests liquid biopsy could be cost-effective if test costs fall to $200 to $300 and sensitivity for advanced precancerous lesions exceeds 70%. However, successful clinical integration also requires physician and patient education, quality assurance programs across laboratories, and follow-up protocols for positive test results, all of which require careful planning and stakeholder engagement.
Several large-scale clinical trials are now underway to establish whether liquid biopsy can replace or complement current CRC screening tools. NCT05244784 is comparing a multi-analyte liquid biopsy panel combining ctDNA mutations, methylation, and proteins against colonoscopy in 50,000 asymptomatic individuals aged 50 to 75, with a five-year follow-up tracking cancer detection rates and interval cancers. Results are expected in 2026 to 2028.
The ECLIPSE study is evaluating a ctDNA methylation test combined with protein markers in 10,000 average-risk individuals, comparing performance to FIT. The PREEMPT CRC study is testing whether liquid biopsy can reduce unnecessary colonoscopies by more precisely triaging individuals who tested positive on FIT. These trials will provide the first rigorous, large-scale evidence on clinical performance in real-world screening populations.
Critical priorities for the next five to ten years include standardizing pre-analytical protocols for blood collection and storage, developing cost-effective testing platforms suitable for population screening, rigorously validating panels in diverse populations across different ethnicities and geographic regions, and establishing regulatory and reimbursement pathways for multi-omics liquid biopsy tests.
The field is progressing rapidly, and the authors project that within a decade, liquid biopsy will complement or partially replace traditional CRC screening tools, enabling detection at earlier stages, personalized risk assessment, and ultimately improved survival for millions of patients worldwide. The technology still faces obstacles, but the scientific foundation is solid and the clinical trials now underway will provide the evidence needed for widespread implementation.