Evolution of non-invasive colorectal neoplasm detection

Abstract

Colorectal cancer (CRC) is one of the leading cancers worldwide in terms of both incidence and mortality. Colonoscopy remains the most accurate and effective method for early detection and screening of CRC, and it also enables CRC prevention by removing adenomas. However, because it is an invasive procedure that requires bowel preparation and carries risks of discomfort and complications, various non-invasive diagnostic tools have been developed as alternatives. These tools use stool, blood, urine, and breath samples, with a particular focus on biomarkers targeting DNA methylation. Currently, strategies that combine multiple biomarkers are under development. Biomarker selection is increasingly guided by machine learning based on next-generation sequencing data. In particular, the concept of multi-omics has played a pivotal role in this development, and many novel diagnostic tools are expected to be validated in large-scale clinical trials. This review aims to enhance our understanding of the principles behind diagnostic tools for early CRC detection by outlining colorectal carcinogenesis and providing an overview of the evolution of CRC screening strategies.

Key Words: Blood; Colorectal neoplasms; DNA methylation; Feces; Screening

Core Tip: Non-invasive colorectal cancer (CRC) screening modalities have been developed as alternatives to colonoscopy to improve patient compliance and reduce procedural risks. DNA methylation biomarkers are the key high-performing component of non-invasive CRC screening tests based on stool and blood. Recent advances in next-generation sequencing and machine learning have enabled the integration of multi-omics data into non-invasive CRC screening, facilitating the development of diverse biomarker-based tests across multiple sample types.

INTRODUCTION

Colorectal cancer (CRC) ranks third in incidence and second in mortality among cancers worldwide, making it a major public health concern[1]. In addition to CRC, lung, liver, stomach, prostate, breast, and cervical cancers are among the most prevalent and deadly cancers globally, in terms of both incidence and mortality, underscoring the critical importance of early detection. While lung and liver cancers are typically screened only in high-risk individuals, such as smokers or those with hepatitis or cirrhosis, and prostate, breast, and cervical cancers are screened in a sex-specific manner, CRC, along with gastric cancer, is screened in the general population. Most CRC develops through the adenoma-carcinoma sequence[2], enabling cancer prevention by detecting and removing advanced adenomas (AAs), defined as lesions ≥ 10 mm in size, containing ≥ 25% villous histology, or exhibiting high-grade dysplasia[3]. In this regard, colonoscopy is the only modality that enables both cancer detection and prevention and is thus recommended as an opportunistic screening tool in many countries[4]. However, colonoscopy involves dietary restrictions, bowel preparation, and often sedation. Its invasive nature and limited accessibility in certain regions contribute to low adherence rates, ultimately limiting its effectiveness as a screening tool for the general population[5]. Consequently, efforts have focused on developing non-invasive diagnostic tools as alternatives, leading to the identification of biomarkers, biological indicators of physiological changes, in stool, blood, urine, and other sample types[6]. This review discusses the principles and current applications of these modalities for diagnosing and predicting CRC and AAs.

COLORECTAL CARCINOGENESIS AND THE ROLE OF EPIGENETIC CHANGES IN SCREENING

Carcinogenesis sequences and classification of CRC

CRC develops through the accumulation of genetic and epigenetic alterations[7]. In 1990, Fearon and Vogelstein[2] established the foundation of the adenoma-carcinoma sequence theory, proposing that CRC progresses from normal colonic epithelial cells through a stepwise accumulation of mutations in oncogenes and tumor suppressor genes, including adenomatous polyposis coli (APC), Kirsten rat sarcoma viral oncogene homolog (KRAS), DCC, and tumor protein 53. Subsequent studies on various molecular and signaling pathways have distinguished not only the traditional intestinal type dysplasia pathway of carcinogenesis, characterized by frequent APC mutations and activation of the Wnt signaling pathway, but also the serrated neoplasia pathway, which is initiated and driven by B-Raf proto-oncogene serine/threonine kinase (BRAF) mutations[8].

CRC is one of the most genetically heterogeneous malignancies, which underlies its diverse molecular classifications and carcinogenic pathways. In 1997, Lengauer et al[9] classified CRCs into two major genetic categories: Tumors with microsatellite instability (MSI), caused by mismatch repair deficiency at the nucleotide level, and tumors without MSI, which exhibit chromosomal instability due to severe defects in chromosome segregation. This classification later proved to broadly correspond to the hypermutated (16%) and non-hypermutated (84%) groups identified in a genome-scale analysis conducted by The Cancer Genome Atlas Network in 2012, based on the burden of DNA mutations per 10⁶ bases[10]. However, as the molecular and signaling pathways underlying colorectal carcinogenesis have been further elucidated, a more comprehensive classification has been proposed, comprising four consensus molecular subtypes (CMS)[11]: CMS1 (MSI immune), characterized by MSI, strong immune activation, and frequent BRAF mutations; CMS2 (canonical), marked by chromosomal instability and activation of the Wnt and Myc signaling pathways; CMS3 (metabolic), featuring frequent KRAS mutations and metabolic dysregulation; and CMS4 (mesenchymal), distinguished by activation of the transforming growth factor-β pathway, stromal invasion, and angiogenesis.

Epigenetic alterations in CRC and their role in screening

Epigenetic alterations in CRC profoundly influence molecular changes and carcinogenic pathways. Among various epigenetic mechanisms, such as chromatin remodeling, histone modification, and RNA-related regulation, DNA methylation is frequently observed in CRC[12]. Compared to the normal colonic epithelium, colorectal neoplasms exhibit global DNA hypomethylation[13], along with regional DNA hypermethylation, most notably in gene promoter regions rich in CpG islands. This focal hypermethylation is a key epigenetic mechanism that silences tumor suppressor genes and contributes to colorectal carcinogenesis[14]. In particular, the CpG island methylator phenotype is closely associated with the MSI phenotype in CRC. In this subtype, hypermethylation of the MutL homolog 1 promoter leads to loss of DNA mismatch repair function, resulting in MSI-high. These changes are also strongly associated with the BRAF V600E mutation and collectively characterize the serrated neoplasia pathway, which originates from sessile serrated adenomas[15]. Therefore, CpG island methylator phenotype is considered a key epigenetic mechanism that not only defines distinct carcinogenic pathways in CRC but also significantly influences the molecular subtypes and clinical behavior of the disease.

Considering the complex molecular and carcinogenic landscape of CRC, considerable efforts have been made to identify actionable biomarkers for the development of non-invasive diagnostic tools. Among the various strategies, epigenetic alterations, particularly DNA methylation, have been preferentially targeted over genetic mutations in biomarker development. This preference is based on several key advantages. Unlike sequentially occurring genetic mutations, DNA methylation events arise at relatively early stages of carcinogenesis, thereby allowing the detection of early-stage colorectal neoplasms[16]. In contrast to genetic mutations, DNA methylation demonstrates high mitotic stability[17] and reflects the tissue-specific epigenetic memory that is clonally maintained over time within individual colonic glands[18,19], rendering it more stable and suitable for biomarker detection. In addition, somatic DNA mutations tend to be heterogeneous and less abundant, and are detected with lower sensitivity than methylation markers[20]. Therefore, studies on non-invasive biomarkers for CRC screening have primarily focused on detecting DNA methylation. To enhance sensitivity, ongoing studies have actively explored the combinations of multiple biomarkers and the development of models based on multi-biomarker panels.

CONVENTIONAL STOOL TESTS

Fecal occult blood test

The fecal occult blood test (FOBT), first introduced as a screening tool for CRC in 1967[21], is based on the principle that the heme component of hemoglobin interacts with guaiac on the test card, initiating a pseudoperoxidase-like reaction that results in a bluish discoloration[22]. This test detects hemoglobin shed into the bowel lumen during bleeding from adenomas or CRC and generally requires the excretion of more than 10 mg of fecal hemoglobin per gram of stool for at least 50% of tests to yield a positive result[23]. In 1996, the United States Preventive Services Task Force officially recommended FOBT, along with other screening tests, for CRC screening in all adults aged 50 and older[24].

However, a limiting factor of FOBT is that bleeding may go undetected even in the presence of CRC or AA if the amount of blood is insufficient. Therefore, stool samples are typically collected over three consecutive bowel movements to increase the likelihood of detection[25]. The oxidation reaction mediated by α-guaiaconic acid, the active compound in guaiac, is not specific to human heme. Consequently, dietary and medication restrictions are required 2-3 days before testing, as certain foods (animal blood-containing meats such as red or rare meat, and peroxidase-rich vegetables including broccoli, carrots, cauliflower, cucumbers, grapefruit, horseradish, melon, mushrooms, radish, raw beets, and turnips) and medications (those causing gastrointestinal bleeding, such as non-steroidal anti-inflammatory drugs, warfarin, heparin, and steroids, as well as substances influencing oxidation, including iron and vitamins C and E) can result in false-positive or false-negative results[26]. However, the effect of these dietary and medication restrictions on CRC detection efficacy has not been thoroughly studied[27].

The effectiveness of FOBT in reducing CRC incidence and mortality has been demonstrated previously. In a meta-analysis of five randomized clinical trials (RCTs) comparing FOBT using Hemoccult II (Beckman Coulter, Brea, CA, United States) with no screening, annual or biennial screening conducted over 11 to 30 years across 2 to 9 screening rounds resulted in reductions in CRC incidence [relative risk (RR): 0.90, 95% confidence interval (CI): 0.77-1.04 to RR: 1.02, 95%CI: 0.93-1.12] and mortality (RR: 0.78, 95%CI: 0.65-0.93 to RR: 0.91, 95%CI: 0.84-0.98)[28]. Following the use of traditional FOBTs, high-sensitivity versions such as Hemoccult SENSA (Beckman Coulter, Brea, CA, United States) were introduced to improve test sensitivity. Nevertheless, their performance, particularly that of Hemoccult Sensa, remains limited, with sensitivity ranging from 0.50-0.75 and specificity ranging from 0.96-0.98[28]. These limitations have driven efforts to develop more specific screening tools for colorectal neoplasms.

Fecal immunochemical test

Fecal immunochemical tests (FITs) were developed in the 1970s to overcome the low specificity of FOBT by using antibodies specific for human hemoglobin[29,30]. In the early stages, the hemoglobin detection threshold of FIT was more than 10 times lower than that of FOBT[30]. A wide range of FIT reagents is commercially available, among which the OC-Sensor (Eiken Chemical Co., Ltd., Tokyo, Japan) is the most widely used[31].

A meta-analysis of 13 studies that used the OC-Sensor with the manufacturer-recommended threshold of 20 μg of hemoglobin per gram of feces demonstrated a sensitivity and specificity for CRC of 0.74 (95%CI: 0.64-0.83) and 0.94 (95%CI: 0.93-0.96), respectively, and 0.23 (95%CI: 0.20-0.25) and 0.96 (95%CI: 0.95-0.97) for AA[28]. As expected, FIT performance varies depending on the positivity cutoff. A meta-analysis of 46 studies evaluating FIT accuracy across different thresholds, stratified by sex and age, showed that lowering the cutoff to ≤ 10 μg hemoglobin per gram of feces increased the sensitivity for detecting both CRC and AA by 11% and 10%, respectively. However, this improvement comes at the cost of increased false positivity[32].

To date, only one large-scale observational study has directly compared FIT with no screening to evaluate its impact on CRC mortality. In this study, which included over 5.4 million individuals, those who underwent biennial FIT screening one to three times showed a significantly lower CRC mortality after six years of follow-up compared to individuals who were not screened (RR: 0.90, 95%CI: 0.84-0.95)[33]. An RCT comparing a single colonoscopy with two rounds of biennial FIT found no significant difference in CRC detection rates[34]. However, another trial comparing single computed tomography colonography with three biennial FITs reported that FIT was not superior at detecting advanced neoplasia[35]. Nevertheless, a recent RCT involving an intention-to-treat population of 26332 individuals found that biennial FIT was non-inferior to one-time colonoscopy in reducing 10-year CRC mortality (RR: 0.92, 95%CI: 0.64-1.32), with participation rates of 39.9% for FIT and 31.8% for colonoscopy, and over two-thirds of FIT participants completing at least 60% of the offered tests[36]. Notably, the United States Preventive Services Task Force currently recommends annual FIT screening for adults aged 45 years and older[37].

Although FIT theoretically does not require dietary or medication restrictions, unlike FOBT, recent evidence has challenged this assumption. One study reported that proton pump inhibitor use was associated with a 39% increased risk of false-positive FIT results[38]. Regarding anticoagulant and antiplatelet agents, most studies reported no significant effect on detection rates[39]. However, one study noted that these medications, along with non-steroidal anti-inflammatory drugs, may slightly increase FIT sensitivity without significantly affecting specificity, suggesting that discontinuation prior to screening is unnecessary[40]. Most importantly, the limited sensitivity of FIT for AA detection remains a key limitation for early detection and cancer prevention. Efforts are ongoing to develop more sensitive diagnostic tools capable of detecting precancerous lesions, particularly by identifying biomarkers specific to colorectal carcinogenesis.

NOVEL STOOL TESTS

Stool DNA mutation tests

Although blood-based DNA tests were developed and studied earlier than stool-based DNA tests[41], stool DNA tests are theoretically more sensitive, as tumor cells undergo luminal exfoliation before vascular invasion occurs during the early phase of carcinogenesis[42]. Initial research on stool DNA tests focused on detecting point mutations in stool, particularly those associated with the adenoma-carcinoma sequence. However, assays targeting APC alone, or multitarget assays combining APC, KRAS, P53, MSI markers, and abnormal apoptosis markers, achieved only modest sensitivity for CRC detection, typically in the low 60% range[43,44]. In one early study, a small-scale trial targeting APC, KRAS, and P53 mutations, Bat-26, and highly-amplifiable or “long” DNA, reported a high sensitivity of 91% for CRC detection. However, due to the small sample size, further evidence is required to validate these findings[45]. ColoAlert (Mainz Biomed Germany GmbH, Mainz, Germany), the only multitarget stool DNA (mt-sDNA) test currently marketed in Europe, combines the detection of KRAS and BRAF mutations, quantification of human DNA, and FIT. However, further studies are needed to strengthen the evidence base[46,47].

Stool DNA methylation tests

Subsequently, research focused on genes known to be methylated in other cancers, such as breast cancer, as well as genes commonly methylated in CRC, beginning with secreted frizzled-related protein 2 (SFRP2)[48]. The representative methylated genes studied include CDN2A, MGMT, TFPI2, N-Myc downstream-regulated gene 4 (NDRG4), syndecan 2 (SDC2), bone morphogenetic protein 3 (BMP3), PRIMA1, septin 9 (SEPT9), SFRP2, and vimentin (VIM), with reported sensitivities and specificities for CRC and adenoma detection varying widely, ranging from 20% to 93.4% and from 80% to 100%, respectively (Table 1)[49].

Table 1 Key stool DNA methylation markers evaluated as single markers for colorectal cancer detection.

Genes	Function	Sensitivity/specificity, %	Ref.
APC	Tumor suppressor gene regulating Wnt signaling and cell adhesion	20-32.1/90-100	[115,116]
BMP3	Tumor suppressor gene regulating transforming growth factor-β pathway and cell differentiation	37.1-84/84-100	[117-121]
CDKN2A	Tumor suppressor gene regulating cell cycle via pRB and p53 pathways	20-40/84-100	[74,122,123]
GATA4	Tumor suppressor gene regulating cell differentiation and transforming growth factor-β signaling	42.9-71.4/84.4-95	[124,125]
HLTF	Tumor suppressor gene regulating DNA repair and chromatin remodeling	20-37.5/30-92.6	[115,126]
hMLH1	Tumor suppressor gene regulating mismatch repair and genomic stability	20-30/86.5-90	[115,127]
MGMT	Tumor suppressor gene reversing DNA alkylation damage	20-51.7/73-100	[74,115,116,123,127,128]
NDRG4	Tumor suppressor gene regulating PI3K/AKT signaling and differentiation	46.7-92/80-100	[117,118,121,125,129,130]
SDC2	Tumor suppressor gene regulating cell adhesion and Wnt signaling	69-95/81-98	[52,53,117,131-135]
SEPT9	Tumor suppressor gene regulating cytokinesis and genome stability	20/80	[136]
SFRP1	Tumor suppressor gene inhibiting Wnt signaling and cell proliferation	29.8-89/86-95.5	[137-139]
SFRP2	Tumor suppressor gene inhibiting Wnt signaling and angiogenesis	30-94.2/70-100	[48,115,117,118,123,125,128,140-144]
TFPI2	Tumor suppressor gene inhibiting extracellular matrix degradation and metastasis	31-92/86.7-100	[118,121,145,146]
VIM	Tumor suppressor gene regulating epithelial–mesenchymal transition and metastasis	41-86/72-100	[121,125-127,136,147-150]

APC: Adenomatous polyposis coli; BMP3: Bone morphogenetic protein 3; CDKN2A: Cyclin dependent kinase inhibitor 2A; GATA4: GATA binding protein 4; HLTF: Helicase like transcription factor; hMLH1: Human MutL homolog 1; MGMT: O6-methylguanine-DNA methyltransferase; NDRG4: N-Myc downstream regulated gene 4; SDC2: Syndecan 2; SEPT9: Septin 9; SFRP: Secreted frizzled related protein; TFPI2: Tissue factor pathway inhibitor 2; VIM: Vimentin; pRB: Retinoblastoma; PI3K: Phosphatidylinositol 3-kinase; AKT: Protein kinase B.

The overall performance of these markers was analyzed in a meta-analysis by Gachabayov et al[50], which included 46 studies on stool DNA methylation tests. The pooled sensitivity and specificity for CRC detection were 0.63 (95%CI: 0.58-0.67) and 0.91 (95%CI: 0.90-0.92), respectively. Notably, assays using gene combinations showed a 23% higher CRC detection sensitivity compared to those using a single gene (0.81, 95%CI: 0.75-0.85 vs 0.58, 95%CI: 0.52-0.63), without a significant decrease in specificity (0.88, 95%CI: 0.84-0.91 vs 0.92, 95%CI: 0.90-0.94). Among the single-gene assays, SDC2 demonstrated the highest accuracy, with a sensitivity of 0.83 (95%CI: 72.6-90.2) and specificity of 0.91 (95%CI: 88.6-93.2).

Early methylation-based stool DNA assays generally focused on single-gene targets. For example, Colosure (Exact Sciences, Madison, WI, United States), which targets methylated VIM, was among the first to be developed[51]. Several SDC2-targeting assays have been developed in Asia and have demonstrated promising clinical performance in prospective studies involving over 1000 participants. For example, EarlyTect (Genomictree, Daejeon, South Korea) reported a sensitivity of 95% for CRC, 58.1% for AA, and a specificity of 81.5% for CRC[52]. In contrast, Colosafe (Creative Biosciences, Guangzhou, China) reported a sensitivity of 83.8% for CRC and 42.1% for AA, with a specificity of 98% in individuals with normal colonoscopy findings[53]. A recent real-world community-based study in China showed higher CRC and AA detection rates in methylated SDC2-positive individuals, although the interpretation was limited by a low colonoscopy completion rate (32.6%)[54]. Several combined methylation assays are currently under validation, such as SpecColon (Suzhou VersaBio Technologies Co. Ltd., Kunshan, Jiangsu, China), which targets methylated SFRP2 and SDC2, and ColoDefense (Suzhou VersaBio Technologies Co. Ltd., Kunshan, Jiangsu, China), which targets methylated SEPT9 and SDC2, although ColoDefense has been primarily utilized as a blood-based assay in most studies[55-59].

Mt-sDNA tests

Early multitarget assays primarily targeting DNA point mutations have shown limited sensitivity for CRC detection. However, in a study conducted by Ahlquist et al[60], the incorporation of VIM methylation into mutation-based panels notably improved test performance, particularly in direct comparisons with FOBT, demonstrating clear advantages over mutation-only assays. This has led to the serial development of mt-sDNA tests that incorporate additional molecular markers. A pivotal study included KRAS mutations, methylated BMP3, NDRG4, TFPI2, and VIM, along with the reference genes β-actin and hemoglobin, quantified via an immunochemical assay[42], which laid the groundwork for the development of subsequent mt-sDNA methylation tests.

Cologuard (Exact Sciences, Madison, WI, United States) is an mt-sDNA test that detects KRAS mutations, aberrantly methylated BMP3 and NDRG4, β-actin, and hemoglobin[61]. The DeeP-C study, conducted by Imperiale et al[61] and published in 2014, was a large cross-sectional trial involving 9989 participants and designed to compare the performance of Cologuard and FIT. The sensitivity of Cologuard was 92.3% for CRC (18.5% higher than FIT) and 42.4% for precancerous lesions, including AA and sessile serrated polyps ≥ 1 cm (18.6% higher than FIT), while the specificity among individuals with no colorectal findings was 86.6% for Cologuard and 94.9% for FIT. A meta-analysis of four RCTs using Cologuard for CRC detection also demonstrated consistently favorable performance[61-64]. Following the publication of the initial study, the United States Food and Drug Administration (FDA) approved Cologuard as the first mt-sDNA test for CRC screening in 2014. Simultaneously, the United States Centers for Medicare and Medicaid Services issued a National Coverage Determination, later finalized in the same year, reimbursing the use of Cologuard every three years for asymptomatic, average-risk beneficiaries aged 50-85 years[65]. A subsequent longitudinal study of approximately 2000 individuals assessed this triennial mt-sDNA screening approach. While no cases of CRC were identified, preventing evaluation of test performance for cancer detection, the study showed significantly enhanced detection of precancerous lesions[66].

In terms of cost-effectiveness, a Markov simulation model study in average-risk individuals reported that, assuming equal adherence, FIT and colonoscopy were more effective and less costly than mt-sDNA screening. For triennial mt-sDNA to be cost-effective, participation rates should be at least 1.7 times higher than those of FIT[67]. A more recent Markov-based study reported favorable cost-effectiveness of mt-sDNA screening; however, the results were based on a specific subpopulation, Alaska native individuals, which limits generalizability to broader populations[68]. Among approximately 360000 Medicare beneficiaries in the United States, comprising individuals aged 65 and older, those with certain disabilities, and patients with end-stage renal disease, nearly three-quarters demonstrated adherence to mt-sDNA testing, indicating relatively high participation[69]. In addition, mt-sDNA has lower specificity than FIT, which may lead to more follow-up colonoscopies, increased adverse events, and higher costs[26]. Amid these concerns, the 2021 guidelines of the American College of Gastroenterology recommended against further evaluation or testing of patients with positive mt-sDNA results but no abnormal colonoscopy findings, due to the risk of false positives[70].

A second-generation test, Cologuard Plus (Exact Sciences, Madison, WI, United States), was approved by the FDA in 2024, with improved specificity and high sensitivity[71]. This updated test incorporates three methylation markers (LASS4, LRRC4, and PPP2R5C), a reference gene (ZDHHC1), and hemoglobin. In validation studies, Cologuard Plus showed a sensitivity of 95% for CRC, 43% for advanced precancerous lesions, and a specificity of 94%, significantly reducing the false-positive rate compared to the original Cologuard[72]. In the 2023 Clear-C trial, the mt-sDNA test ColoClear (New Horizon Health, Hangzhou, Zhejiang Province, China), which targets KRAS mutations, methylated BMP3 and NDRG4, β-actin, β2-microglobulin, and hemoglobin, demonstrated a sensitivity of 91.9% for CRC, 63.5% for AA, and a specificity of 87.1%-90.3% in a cohort of 4245 participants[73]. However, as the study population included high-risk individuals (e.g., those with a family history, prior positive stool tests, or symptoms), the findings may not be directly generalizable to average-risk populations. These developments reflect a broader trend toward combining multiple biomarkers through algorithm-based modeling to optimize CRC detection, an approach that is also being explored in emerging blood-based tests[74].

Stool RNA tests

In addition to detecting CRC-specific DNA in stool, research has advanced toward targeting messenger RNA (mRNA) transcribed from it. Barnell et al[75] initially identified 15 significant mRNA transcripts detectable in stool through reproducibility testing based on sampling. In a subsequent study, this was narrowed down to eight targets (ACY1, AREG, CDH1, EGLN2, GAPDH, KRAS, SMAD4, and TNFRSF10B). These results were combined with FIT results and participant-reported smoking status to develop a multitarget stool RNA test, ColoSense (Geneoscopy Inc., St. Louis, MO, United States)[76]. In the CRC-PREVENT trial published in 2023, which evaluated 8920 average-risk individuals aged 45 years or older, the test demonstrated significantly higher performance compared to FIT, with a sensitivity of 94% for CRC, 46% for AA, and a specificity of 88%, and was later approved by the FDA[77].

Non-coding RNAs, which comprise the majority of the RNA in the human genome and are not translated into proteins, play essential roles in regulating gene expression, cell growth, carcinogenesis, and metastasis[78]. MicroRNAs (miRNAs) primarily function by degrading mRNA or inhibiting its translation, and can act as either oncogenes or tumor suppressors. Since the study by Link et al[79] in 2010, research on miRNAs was primarily conducted in experimental settings, showing a wide range of sensitivity (34.2%-97%) and specificity (48%-100%). However, large-scale studies and standardization are still lacking for both stool- and blood-based tests. Challenges such as ethnic differences in miRNA expression and a lack of organ specificity highlight the need for future studies to improve the accuracy and reproducibility of miRNA-based diagnostics[80].

Stool microbiome tests

Among the substances detected in stools, the gut microbiome is known to play an important role in colorectal carcinogenesis and CRC progression[81]. Among these microbes, Fusobacterium nucleatum (F. nucleatum) is the most actively studied[82]. DNA metagenomic sequencing of gut bacteria, including F. nucleatum, often combined with machine learning, has primarily been used to develop diagnostic tools for colorectal neoplasms[83,84]. Notably, in 2021, Cao et al[85] reported that the combination of the integrin subunit alpha 4 methylation marker, F. nucleatum, and Peptostreptococcus anaerobius, when used alongside FIT, achieved high diagnostic performance with sensitivities of 95.4% for CRC and 70.8% for AA, and it holds promise pending further validation in real-world settings.

Stool inflammatory marker tests

Among the substances shed into the stool by colorectal neoplasms, several studies have evaluated the diagnostic potential of inflammatory markers in addition to cellular components. Representative proteins include M2-pyruvate kinase and fecal calprotectin, both of which are known to correlate with intestinal inflammation in inflammatory bowel disease[86,87]. However, a meta-analysis of 65 studies on FIT, M2-pyruvate kinase, and fecal calprotectin reported that the detection of colorectal neoplasms using fecal proteins was clinically viable (areas under the curve > 0.7) only when combined with FIT[88].

BLOOD TESTS

Liquid biopsy

Historically, efforts to identify diseases using substances (biomarkers) derived from the human body have often been first explored through blood tests. Howard Scher is credited with introducing the term “liquid biopsy” in the context of research on circulating tumor cells (CTCs)[89]. Although the term may refer to a variety of biological fluids such as urine, sputum, peritoneal fluid, pleural fluid, and cerebrospinal fluid, it predominantly denotes a sampling test performed using peripheral blood[90]. This test targets somatic components in blood, including CTCs (rare cancer cells shed into the bloodstream), circulating nucleic acids, and extracellular vesicles.

CTCs

In 1869, Ashworth[91] observed blood cells resembling tumor cells during the autopsy of a deceased patient, marking the initial description of CTCs. However, due to the extremely low concentration of CTCs in the blood, it was only after significant advancements in diagnostic technologies that associations between CTCs and cancer, as well as cancer prediction, began to be established[41]. The utility of CTCs for diagnostic or screening purposes has remained limited, as even in 1 mL of blood, the number of CTCs is exceedingly low, heterogeneous, and lacks cancer-specific markers[41]. In particular, CTC detection rates tend to increase with advancing cancer stage, 4.9%, 10.5%, 8.3%, and 18.8% for Union for International Cancer Control stage I, II, III, and IV, respectively[92], and CTC positivity is low in early-stage CRC[93]. As such, CTCs are better known for their prognostic rather than diagnostic value in CRC[94].

In 2019, Tsai et al[95] developed a predictive algorithm for the presence of CRC or adenomas using CTC counts from blood samples collected from 667 participants prior to colonoscopy. Their model demonstrated 95% sensitivity for CRC across all stages, 71.4% for AA, and 86% specificity for individuals with no findings. Notably, the sensitivity for stage I CRC was also high (89.2%), suggesting strong detection performance. These results support the screening potential of CTC detection and highlight the need for further long-term studies.

Circulating tumor DNA tests

In 1948, Mandel and Metais[96] reported the presence of nucleic acids in the blood, and in 1989, Stroun et al[97] reported the presence of tumor-derived cell-free DNA (cfDNA) in the plasma of patients with cancer, which later led to the concept of circulating tumor DNA (ctDNA). Similar to stool tests, CRC screening tests using ctDNA have primarily been developed to detect DNA methylation rather than DNA mutations. The target genes included SEPT9, SHOX2, BCAT1, IZFF1, SFRP1, SFRP2, SDDC2, PRIMA1, and VIM[90].

The first blood test approved by the FDA for CRC screening was Epi proColon (Epigenomics AG, Berlin, Germany), which detects SEPT9 methylation in ctDNA. In the 2014 PRESEPT trial involving 7941 participants, first-generation Epi proColon showed a sensitivity of 48% and a specificity of 92% for CRC[98]. However, its sensitivities for stage I CRC and AA were only 35% and 11%, respectively, raising concerns about its effectiveness in preventing early-stage neoplasms. Subsequently, a second-generation version of Epi proColon was developed, which was considered positive if even one of the three polymerase chain reactions was positive, unlike the first-generation version, which required all three polymerase chain reaction reactions to be positive[99]. A meta-analysis published in 2020 analyzing 19 studies on blood tests targeting SEPT9 methylation reported a pooled sensitivity and specificity of 69% (95%CI: 62-75) and 92% (95%CI: 89-95), respectively, suggesting that the performance may not be highly promising[100]; however, this conclusion is limited by substantial heterogeneity across studies (I² = 88.4% for sensitivity and 96.3% for specificity). Notably, the FDA-approved Epi proColon is indicated only for average-risk individuals aged 50 or older who have been offered screening tests but declined or failed to undergo them[101].

ColoVantage (Quest Diagnostics, Secaucus, NJ, United States), a blood-based test targeting SEPT9 methylation, was developed for commercial use; however, it has not yet been clinically validated as a screening test[102]. This test differs from ColoVantage Home (Clinical Genomics Pty Ltd., Sydney, Australia), an FIT-based product, in its application and underlying mechanism[103]. Several other commercially available blood-based tests have been developed, including ColonAiQ and ColoScape. ColonAiQ targets methylation of SEPT9, BCAT1, IKZF1, BCAN, and VAV3, whereas ColoScape detects mutations in eight genes (APC, KRAS, BRAF, tumor protein 53, CTNNB1, NRAS, SMAD4, and PIK3CA) and seven methylation markers[104,105]. The reported sensitivity and specificity were 86% and 92%, respectively, for ColonAiQ, and 86% and 100%, respectively, for ColoScape. These results are based on retrospective studies; prospective, large-scale validation is required.

The analysis of cfDNA fragment profiles, such as fragment size, motifs, and fragment ends, is termed fragmentomics[106]. Recently, this concept has been integrated with genomics and epigenomics under a broader multi-omics framework to develop CRC screening tools using liquid biopsies. The ECLIPSE trial (2024) and the interim analysis of the PREEMPT CRC trial (2025) evaluated CRC detection tools named Shield (Guardant Health Inc., Redwood City, CA, United States) and Freenome (Freenome Holdings, Inc., South San Francisco, CA, United States) using next-generation sequencing and machine-learning approaches on cfDNA in cohorts of 22877 and 48995 participants, respectively[107,108]. Both studies reported good overall sensitivity for CRC (83.1% and 81.1%, respectively), but the sensitivities for stage I CRC and precancerous lesions were 65% and 57.1%, and 13.2% and 13.7%, respectively. Thus, both platforms appear promising as complementary screening tools, particularly for individuals who are unwilling to undergo a colonoscopy.

There has been ongoing research on multi-cancer early detection (MCED) tests, which aim to detect various cancer types at early stages using combinations of biomarkers[107]. Two representative studies in this field are the DETECT-A study using CancerSEEK (Exact Sciences, Madison, WI, United States), which is based on the multi-omics analysis of ctDNA, and the PATHFINDER study using Galleri (GRAIL, LLC, Menlo Park, CA, United States), which detects cancer signals through methylation patterns in cfDNA. In the DETECT-A study (2020), MCED testing in 9911 cancer-free women aged 65-75 identified 26 new cancers (sensitivity: 27%); among three CRC cases, two (stage II and III) were detected, while one stage I case was missed[109]. In the PATHFINDER study, among 6621 asymptomatic adults aged ≥ 50, cancer signals were detected in 92 (1.4%), of which 35 were true positives (38%). For CRC, two stage IV cases were detected, while one stage I case was missed[110]. Currently, several large-scale clinical trials using MCED tests are ongoing, including the ASCEND, PATHFINDER-2, NHS-Galleri, PREVENT, and Vanguard studies[111].

Other blood tests

Numerous serum miRNA tests have been studied as tools for CRC screening, with reported sensitivities ranging from 58.9% to 93.1% and specificities ranging from 72.7% to 89.1%[80]. A meta-analysis of 20 studies reported a pooled sensitivity of 0.85 (95%CI: 0.84-0.86) and specificity of 0.79 (95%CI: 0.78-0.80)[112]. However, this analysis also noted high heterogeneity across studies (I² for sensitivity and specificity: 96.3% and 98.3%, respectively), suggesting that firm conclusions about miRNA performance are not yet justified. Currently, miRNA-based tests have not received official approval for CRC screening. The limitations of this test include low organ specificity and variability in miRNA expression across ethnic groups, indicating that further research is needed[80].

OTHER TESTS

In addition to the CRC screening methods using blood and stool samples described above, diagnostic tools using urine and breath samples have also been explored. In particular, volatile organic compounds (VOCs), carbon-based molecules, have been investigated based on the premise that metabolic byproducts differ between cancerous and normal tissues. Rather than a single VOC being specific to CRC, the altered pattern of VOCs is thought to be associated with the disease state[113]. A meta-analysis of 15 studies using VOCs derived from stool, urine, and breath reported a sensitivity of 0.84 (95%CI: 0.78-0.88) and a specificity of 0.80 (95%CI: 0.71-0.87) for CRC detection[114]. However, these studies have not yet undergone large-scale, population-based validation, and limitations remain due to heterogeneity in data collection and analysis methods[82].

CONCLUSION

A wide range of non-invasive diagnostic tools for CRC screening has been developed, including stool-based tests and blood-based assays. Conventional methods such as FOBT and FIT have demonstrated their ability to reduce CRC incidence and mortality. More recently developed tools that offer higher detection rates require longitudinal validation in large-scale clinical studies. Blood-based tests are being actively developed as promising alternatives for individuals with low adherence to, or those who do not participate in, conventional screening programs, despite variability in detection accuracy across cancer stages and platforms. Future efforts should focus on integrating multi-omics data with machine-learning approaches to further enhance the accuracy and clinical utility of CRC screening.