Due to its convenience and non-invasiveness, the screening of blood markers is associated with better compliance compared to the other two methods. Our search yielded 11 articles encompassing 21 genes that utilized serum or plasma samples for EC early detection. The outcomes of these studies are summarized in Table 3.

Daito et al. conducted a study in 2001 using serum methylated P16 to detect early EC, but achieved a low sensitivity of 22.6% (Hibi et al., 2001). In another study by Jin et al., 2007, the evaluation of TAC1 methylation for EAC was performed on both plasma and tissue samples simultaneously. The results indicated that TAC1 exhibited similar specificity in plasma and tissue (91.4% vs. 92.5%), but the sensitivity significantly decreased (29.5% vs. 61.2%). Similar findings were observed in Yan et al.'s study, where the sensitivity of methylated PTPRO was 75.0% in tissue but only 36.1% in plasma (You et al., 2012). The lower sensitivity of methylation markers in blood compared to tissue is attributed to the lower abundance of circulating tumor DNA (ctDNA) in the blood. Nevertheless, there are still individual gene markers, such as MGMT reported by Das et al., 2014, which demonstrate good sensitivity of 70.0%. However, employing a marker panel appears to be a preferable approach for blood-based methylation screening. Li et al., 2011 evaluated the detection efficiency of 5 methylated genes, including P16, DAPK, RAR-β, CDH1, and RASSF1A, in serum in 2011. When considering single gene detection, the sensitivity of these 5 genes was 6.7%, 73.3%, 26.7%, 84.4%, and 62.2%, respectively, with corresponding specificities of 100%, 86.7%, 86.7%, 80.0%, and 93.9%. However, when these 5 genes were combined, the sensitivity increased to 82.8% with a specificity of 100%. Qin et al. identified 23 candidate methylation markers from tissue samples that exhibited sensitivity for both EAC and ESCC. Subsequently, they selected 12 methylation markers for plasma testing and narrowed down to 5 markers (FER1L4, ZNF671, ST8SIA1, TBX15, ARHGEF4) to develop a panel for detecting both EAC and ESCC. This panel demonstrated sensitivities of 74% for EAC and 78% for ESCC, with a specificity of 91% (Qin et al., 2019). However, the sensitivity of the 5-gene panel in detecting stage I EC was only 43% (Qin et al., 2019). Bian et al. selected 2 markers (KCNA3 and OTOP2) from 5 methylation markers and validated them in both the training and validation sets. The 2-marker panel demonstrated good diagnostic performance for ESCC in both the training and validation sets, with AUCs of 0.91 and 0.88, respectively. Additionally, it showed a sensitivity of 78.4% for stage I-II ESCC (Bian et al., 2023). Pei et al. developed a 2-marker panel using ZNF582 and FAM19A4 in 0.5–1 mL of plasma, but the overall sensitivity and specificity were not high (Table 3) (Pei et al., 2023).

In addition to the commonly used MSP and qMSP, mass spectrometry and NGS have also been utilized in the detection of EC in blood samples. For instance, Wang et al. constructed a 3-gene panel (CASZ1, CDH13, ING2) using matrix-assisted laser desorption/ionization time-of-flight mass spectrometry and validated it with 0.2 mL plasma samples, achieving an overall AUC of 1.0 (Wang et al., 2018). However, this study only included 10 cases of ESCC and 3 control cases, necessitating more samples to further evaluate the technology (Wang et al., 2018). On the other hand, Qiao et al. identified 921 differentially methylated regions based on tissue samples and constructed a plasma diagnostic model for EC by using deep targeted bisulfite sequencing. They tested the model in three independent cohorts and achieved good sensitivities (74.7%–86.0%) and specificities (94.0%–95.9%) (Qiao et al., 2021). However, the sensitivity of this diagnostic model for stage 0-II esophageal cancer was only 58.8% (Qiao et al., 2021).

Additionally, it is worth mentioning that SEPT9, a widely used detection marker for colorectal cancer (CRC) (Zhao et al., 2019; Zhang et al., 2021), exhibited promising specificity of 92.6% for EC detection (Zhang et al., 2022).

The detection of multiple cancer types through the use of individual methylated genes or a panel of methylated genes, referred to as a pan-cancer test, represents a novel approach aimed at reducing cancer morbidity and mortality (Duffy et al., 2021). This review provides a summary of seven pan-cancer tests utilizing methylation markers, which have been applied to at least two cancer types, including EC (Table 4).

In the detection of ESCC and GC, Liu et al., 2015 reported a sensitivity of 31.0% for EC and a specificity of 88.1% using methylated SFRP1 in serum. Similarly, Song et al., 2020 used methylated SEPT9 in plasma for the detection of four cancers (EC, gastric cancer [GC], hepatocellular carcinoma [HCC], and CRC) and achieved a higher sensitivity and specificity of 42.6% for EC and 94.6% respectively. Peng et al., 2022 developed a panel combining ZNF582, ELMO1, and TFPI2, which allowed for the simultaneous detection of GC, EC, and esophagogastric junction cancer (EJC), with a sensitivity of 79.2% for EC and a specificity of 90.0%. In 2023, Dai et al., 2023 developed a 6-marker panel (KCNQ5, C9orf50, CLIP4, ELMO1, ZNF582 and TFPI2) to detect of EC, GC and CRC, it achieved sensitivities for detecting EC of 64.7% and 83.35 in training and validation sets with specificities of 94.1% and 86.7%.

In 2020, GRAIL, Inc. published a novel multi-cancer detection panel consisting of 1,116,720 CpG sites by using the cfDNA in 10 mL plasma, for detecting 12 types of cancer, including EC, head and neck cancer, CRC, and lung cancer, they validated this panel in two large cohorts and achieved 82.0% and 81.0% sensitivities in training and validation sets, with specificities of 99.8% and 99.3%, respectively, while the sensitivity for stage I EC was only 16.7% (Liu et al., 2020). Next year, Klein et al. optimized this panel and also detecting 12 types of cancer by using a panel of over 100,000 methylation regions in plasma. They obtained a sensitivity of 85.0% for EC and a specificity of 99.5%, but the sensitivity for stage I EC was as less as 12.5% (Klein et al., 2021). It is worth noting that, despite the introduction of the cancer signal origin function in this panel, it is still unable to distinguish between EC and GC (Liu et al., 2020; Klein et al., 2021). Kandimalla et al. developed a targeted bisulfite sequencing-based panel for detecting five types of cancer, utilizing a reduced volume of plasma (Kandimalla et al., 2021). They achieved impressive AUC values of 0.94 for ESCC and 0.90 for EAC. However, it is worth noting that they did not assess the performance of the panel in detecting early-stage ESCC and EAC (Kandimalla et al., 2021). Furthermore, this panel demonstrated high accuracy in distinguishing between ESCC/EAC and other digestive tract cancers (Kandimalla et al., 2021). In 2023, the data of a large clinical trial (called The THUNDER study) for a customized panel with 161,984 CpG site for detecting six types of cancers was published, it can detect 59.5%–80.0% EC in three cohorts with super high specificities, but the sensitivities stage I EC still relatively low (Gao et al., 2023).

EC is a highly lethal disease associated with a poor prognosis, emphasizing the importance of early screening to improve patient survival rates and quality of life. DNA methylation, a widely studied epigenetic modification, is considered a promising tool for cancer screening due to its common occurrence, early onset, and stability during tumorigenesis (Ruddon, 2010). It can be detected in various sample types, including tissues, exfoliated cells, and body fluids such as blood, stool, urine, and cerebrospinal fluid (Li et al., 2014; De Mattos-Arruda et al., 2015; Wang et al., 2015; Husain et al., 2017; Moinova et al., 2018). In this review, we summarized the literature pertaining to DNA methylation detection in EC with different sample types, and assessed the potential and challenges of using DNA methylation as an early detection/screening tool for EC. For those sample types, DNA methylation in tissue is no a suitable sample for EC early detection, because it is an invasive sample, which mostly be used for pathological diagnosis and the discovery stage of DNA methylation markers (Table 5). While esophageal exfoliated cells and blood are two recommended sample types for EC early detection, although the sensitivity and specificity of DNA methylation marker in blood are lower than those in esophageal exfoliated cells, but the high compliance of blood will increase participation rate in early screening of EC (Table 5).

Theoretically, the methylation level in cancer tissue samples should be higher compared to other body fluid samples, such as blood, as suggested by previous studies. This is because the ctDNA in the blood originates from apoptotic cancer tissue, and this only accounts for a small portion of the cancer tissue. After entering the bloodstream, ctDNA undergoes systemic dilution, resulting in lower concentrations in blood. In the extracted blood cfDNA, only a small fraction is ctDNA, with the majority derived from normal cells (Thierry et al., 2016). Therefore, most of the reviewed literature in this study supports this observation. However, there may be exceptions in individual studies. Li et al. evaluated the performance of detecting EC using five methylated genes (P16, DAPK, RAR-β, CDH1, RASSF1A) in both tissue and blood samples. They found that only P16 and RAR-β exhibited higher sensitivity in tissues compared to blood, while the other three genes showed the opposite trend (Li et al., 2011).

The accuracy of the DNA methylation test can be influenced by the selection of target CpG sites and the design of the panel. It is crucial to ensure that the chosen CpG sites are informative and specific for the target cancer types, and that the panel design is optimized for sensitivity and specificity. In addition, significant differences were observed when comparing the detection ability of the same gene in tissue across two different studies. As shown in Table 1, in Zou et al., 2005 study, the sensitivity and specificity of SFRP1 were reported as 92.5% and 90%, respectively. However, in the study conducted by Meng et al., 2011 the sensitivity and specificity of SFRP1 were 95.0% and 35.0%. It is worth noting that the experimental group in Zou et al.'s study consisted of patients with EAC and BE, whereas in Meng et al.'s study, the experimental group comprised patients with ESCC. EAC and ESCC are two completely different types of cancer in terms of molecular subtyping. The clinical treatment strategies for these two cancers are also completely different (Cancer Genome Atlas Research NetworkAnalysis Working Group: Asan UniversityBC Cancer AgencyBrigham and Women’s HospitalBroad InstituteBrown University et al., 2017). Therefore, it is not surprising that the observed significant differences in SFRP1 methylation in EAC and ESCC were found in the aforementioned studies. When comparing the performance of the same marker, it is essential to consider whether the enrolled subjects are consistent. Different subtypes of EC represent distinct origins, resulting in varying sensitivities even when examining the same gene and methylation site. Additionally, the selection of the control group is of great importance. Zou et al.'s study employed normal squamous (SQ) esophageal tissue as the control group, while Meng et al. used para-carcinoma tissue. This selection significantly impacts the specificity of detection. It is observed that many studies did not clearly define the scope of para-carcinoma tissue, making it challenging to establish a uniform comparison across articles and differentiate between para-carcinoma tissue and SQ. Consequently, the lack of consistency in control group selection directly affects the longitudinal comparison of results.

Esophageal balloon is an emerging diagnostic device for EC. This method utilizes capsule sponge-on-string devices specifically designed to capture cells from the esophageal mucosa, offering a direct sampling of the affected tissue. Patients swallow these esophageal sampling devices, which consist of a sponge attached to a string. After a few minutes, the device is retrieved through the mouth, and the sponge is then examined for the presence of cancer cells or DNA (Iyer et al., 2020). Compared to traditional endoscopy, this technique inherits the advantages of providing a direct sample of the affected tissue, thereby potentially increasing sensitivity and specificity while minimizing invasiveness and improving patient compliance. Moreover, esophageal sampling devices can collect millions of cells (Wang et al., 2019), resulting in higher sensitivity compared to cfDNA-based method, especially in detecting early-stage EC. Meanwhile, since the samples collected by esophageal sampling devices are exclusively from the esophagus, they avoid interference from other organs during DNA methylation analysis, resulting in high specificity (Figure 1). For instance, DNA methylation markers such as TFPI2, NDRG4, and BMP3 have been identified as effective markers for early detection/screening of CRC in blood or stool (Imperiale et al., 2014; Rasmussen et al., 2016; Rokni et al., 2018). However, when detecting these markers in esophageal exfoliated cells collected using esophageal sampling devices (Table 2), we can confidently attribute these methylation signals to the esophagus rather than the colon. Furthermore, it is important to note that esophageal sampling devices are more invasive compared to cfDNA-based methods, necessitating specialized equipment and trained personnel for the procedure. Additionally, the risk of complications such as bleeding, mucosal injury or perforation cannot be ignored (Figure 1) (Iyer et al., 2018; Januszewicz et al., 2019). Therefore, the utilization of esophageal sampling devices still has an approximate noncompliance rate of 10% during the application (Table 2). However, current DNA methylation studies based on the esophageal exfoliated cells mainly focus on EAC and its precursor lesions, with only one study specifically targeting ESCC, and a lack of validation for ESCC precursor lesions (Ma et al., 2022). ESCC constitutes the majority of EC cases (Yang et al., 2020), making it essential to pay more attention to the methylation analysis using esophageal exfoliated cells in ESCC in future research. This would provide a feasible pathway for early prevention of ESCC.

Blood testing, as a non-invasive method, offers high compliance. Compared to other approaches, the collection and processing of blood is a routine procedure, and DNA methylation analysis can be carried out using standard laboratory techniques with high-throughput, which may be more cost-effective than obtaining esophageal exfoliated cells (Figure 1) (Wang et al., 2021). Blood testing not only serves as a screening tool for cancer but also allows for more frequent and safer monitoring of response to anticancer therapies in clinical practice (Rothwell et al., 2019). However, cfDNA-based methods have a notable drawback, lacking specificity due to the interference from other organs. They can detect DNA fragments released from non-cancerous cells or other cancer tissues, leading to false positives. Furthermore, the sensitivity of DNA methylation markers in blood, particularly in early-stage disease, regardless of the detection approach, is relatively low (Table 3, 4). This limitation hinders their clinical utility (Figure 1).

Plasma is more commonly utilized than serum, likely due to the lower fraction of ctDNA in serum compared to plasma, as well as higher background noise and larger DNA fragments, as supported by numerous previous studies (Lee et al., 2020; Pittella-Silva et al., 2020). Moreover, there is considerable variation in plasma volumes reported in the literature, ranging from 0.2 to 3.5 mL, with some studies lacking sufficient explanation (Table 3, 4). This disparity in plasma volume can significantly impact sensitivity comparisons. Insufficient plasma volume may result in lower cfDNA yield and poorer quality, ultimately compromising the sensitivity and specificity of methylation analysis. Therefore, in clinical applications, it is often necessary to draw a larger amount of blood to improve sensitivity. However, excessive blood collection may cause discomfort or reluctance among participants. Thus, current cfDNA methylation testing typically recommends drawing 10 mL of blood, from which 3–4 mL of plasma is separated for subsequent analysis.

Throughout the literature reviewed in this review, ZNF582 and TFPI2 emerged as highly promising DNA methylation markers for early detection of EC. ZNF582 was reported twice in tissue samples (Table 2), with both studies indicating good sensitivity and specificity (Huang et al., 2017; Li et al., 2019). Additionally, a study using plasma samples showed ZNF582s favorable sensitivity in detecting EC (Peng et al., 2022; Bian et al., 2023; Dai et al., 2023; Pei et al., 2023). Similarly, TFPI2 has been validated in tissue samples (Jia et al., 2012), esophageal exfoliated cells (Chettouh et al., 2018), and plasma samples (Peng et al., 2022; Dai et al., 2023) from EC patients. However, both ZNF582 and TFPI2 face a common challenge: they are not specific to EC as methylation markers. For instance, ZNF582 exhibits high methylation levels in GC and cervical cancers (Li et al., 2019; Peng et al., 2022), while TFPI2 displays elevated methylation in GC and CRC (Hibi et al., 2011; Peng et al., 2022). When detecting these markers in esophageal epithelial exfoliated cells, interference from other organs is minimized. Yet, when using blood as the testing sample, interference from other organs may result in false positives. In fact, most blood DNA methylation markers are considered pan-cancer markers. For instance, SEPT9 shows methylation positivity in plasma samples from CRC, GC, EC, HCC, and cervical cancers (Potter et al., 2014; Oussalah et al., 2018; Cao et al., 2020; Zhang et al., 2022; Bu et al., 2023), while P16 also serves as a common pan-cancer methylation marker (Hibi et al., 2001; Zou et al., 2002; Hou et al., 2005; Lou-Qian et al., 2013). Therefore, detecting DNA methylation markers in EC blood samples often requires specific population screening or the use of tissue-origin algorithms to avoid false positive signals. Meanwhile, in blood sample testing, the future trend will likely involve the combination of multiple DNA methylation markers to enhance sensitivity for early-stage cancer detection.

In fact, apart from sample volume, there are several other sample preprocessing steps that significantly impact the performance of blood DNA methylation tests, such as the use of a preservation solution before freezing or centrifugal treatment after plasma collection can play a role (Kerachian et al., 2021). Additionally, the storage temperature of samples (ranging from −80°C to 4°C) and the time interval between sample collection and cryopreservation (ranging from 30 min to 24 h or even until the sample is tested) can also influence the results (Kerachian et al., 2021). The extraction and conversion methods of DNA, as described in the literature (whether using self-made reagents or commercialized kits), can affect the amount of DNA extracted, thereby potentially impacting the detection sensitivity and specificity. Studies have demonstrated notable differences in cfDNA recovery efficiency and bisulfite conversion efficiency among various cfDNA isolation kits and bisulfite conversion kits (Sorber et al., 2017; Worm Ørntoft et al., 2017). Therefore, standardizing the operating procedures and implementing quality control measures are crucial to ensure accurate and reliable test results.

MSP, qMSP, and bisulfite NGS are three commonly used methods for methylation analysis. MSP, a traditional method, has been widely employed in various sample types due to its ease of use and low cost. However, its sensitivity and specificity are limited by the potential for cross-contamination and the inability to detect low-frequency methylated DNA (Mao and Chou, 2010; Ramalho-Carvalho et al., 2018). Moreover, MSP lacks the capability to quantitatively assess markers, which is a significant drawback. On the other hand, qMSP, a modified method combining MSP and qPCR, allows for quantitative analysis while minimizing cross-contamination. Nevertheless, it is also limited to a small number of CpG sites and has difficulty detecting low-frequency methylated DNA (Sigalotti et al., 2019). Currently, several FDA and Chinese National Medical Products Administration (NMPA) approved non-invasive cancer early detection tests are based on qMSP, which are valued for their cost-effectiveness and convenience (Potter et al., 2014; Wu et al., 2016; Wang et al., 2020). Notably, the commercially available Epi proColon kit, which examines SEPT9 methylation in blood, has achieved significant success in cancer detection (Potter et al., 2014; Song and Li, 2015). In contrast, bisulfite NGS, a high-throughput analytical method for DNA methylation markers, has not demonstrated a significant advantage over qMSP-based approaches. Its high cost and complex operational process have limited its widespread application (Luo et al., 2020).

Single cancer tests are specifically designed to detect a particular type of cancer, such as CRC in the case of the Epi Procolon test. However, they are not intended to detect other types of cancer, which can be a limitation when a patient has a different cancer type or multiple cancers. Additionally, the cost of single cancer tests can be prohibitive, posing a barrier to access for patients who cannot afford them or for healthcare systems with limited resources. In contrast, pan-cancer tests have the ability to detect multiple cancers in a single tube reaction, offering the potential to revolutionize cancer detection and treatment. This review summarizes several studies on pan-cancer detection, and Table 4 demonstrates their noteworthy performance. This approach presents a novel concept for future cancer screening in specific systems, such as gastrointestinal cancers or gynecologic cancers. However, how to enhance the sensitivity of pan-cancer tests for early-stage cancer diagnosis and effectively reduce the testing cost remains a challenge that needs to be addressed in future research.

Based on the different sample types and analytical methods mentioned above, we summarized a flowchart suitable for developing and validating a DNA methylation assay for early EC detection (Figure 2). In the initial phase of marker discovery (Phase I), it is advisable to utilize FFT samples instead of FFPE samples to mitigate potential DNA degradation and loss, thus minimizing information loss. This stage necessitates the inclusion of diverse sample and disease types, encompassing a comprehensive range of EC samples across various stages, while ensuring age consistency between the EC and control groups, to identify the most specific candidates. NGS stands out as the optimal method for marker discovery due to its high throughput and potential for novel markers.

During the subsequent phase of marker selection (Phase II), it is advisable to validate the initially selected overlapping candidates across additional tissue cohort or databases to mitigate candidate preference. Subsequently, candidate validation should be performed using blood or esophageal exfoliated cells, as methylation levels in tissues may not entirely correlate with those in plasma or esophageal exfoliated cells, especially in plasma samples. Consistent with tissue validation, age, disease type, cancer stage, and cancer location distributions in plasma or esophageal exfoliated cell cohorts should be uniform. Concurrently, maintaining consistency in sample processing methods throughout assay development is essential. Despite the abundance of commercial DNA extraction and methylation conversion kits, thorough validation across multiple batches is imperative to ensure result robustness. Regarding plasma samples, sampling is typically conducted using EDTA tubes, which should be stored at room temperature for a maximum of 4 h or at 2°C–8°C for no more than 24 h, with plasma separation completed within the latter timeframe. Currently, cfDNA collection tubes permit blood samples to be stored at room temperature for up to 7 days before plasma separation, thereby enhancing the convenience of plasma-based detection methods (Hidestrand et al., 2012). Esophageal exfoliated cells are typically preserved in liquid-based cytology medium, such as PreservCyt medium (Hologic, Inc., Marlborough, MA, United States), and can be stored at room temperature for up to 1 month.

Subsequently, the remaining candidates are utilized in constructing the diagnostic model (Phase III). This step necessitates incorporating an adequate number of early EC and control samples, along with precancerous lesion and interfering samples, to derive relatively accurate diagnostic models and cut-off values. While combining more markers often yields higher sensitivity, in light of comprehensive costs and routine screening practices’ accessibility, we advocate employing multiplex qMSP methods for model construction. Following the establishment of the diagnostic model, it is imperative to validate its accuracy and repeatability once more across an adequate number of validation cohorts. It is worth noting that, during multicenter validation, efforts should be made to have some geographical diversity among the centers. For example, a three-center validation could be distributed across East, South, and North China. At the same time, attention should be paid to factors such as the race and dietary habits of the recruited population. Lastly, revalidating the diagnostic model’s accuracy in prospective samples is imperative, along with recommended validation in real-world populations (Phase IV). Nevertheless, validation in real-world populations frequently poses challenges, including population diversity, analysis of confounding factors, and financial support.

Cancer initiation and progression are regulated by a combination of genetic and epigenetic events. The complexity of carcinogenesis extends beyond genetic mutations alone and encompass epigenetic modifications as well (Kanwal and Gupta, 2012). Epigenetics is formally characterized as heritable alterations in gene expression or chromosomal stability through mechanisms such as DNA methylation, histone modifications, or non-coding RNAs (e.g., miRNA) without a change in DNA sequence (Ilango et al., 2020). Therefore, in addition to DNA methylation markers, there are currently several other epigenetic and protein markers being explored for the early detection of EC. These include traditional blood tumor markers (such as CEA, Cyfra21-1, p53, SCC-Ag and VEGF-C) (Zhang et al., 2015), DNA fragments (Tomita et al., 2007), mRNA (Kashyap et al., 2009), and miRNA (Xue et al., 2024). However, most marker studies currently lack in-depth investigation and are deficient in repetitive validation. Studies on DNA methylation and miRNA are the most extensive. For example, Jinsei et, al developed an 8-miRNA panel for early detection of ESCC, and verified in multiple cohorts with AUC values of 0.80–0.93 (Miyoshi et al., 2022). Kazuki et, al developed a 6-miRNA panel with sensitivity and specificity of 96% and 98% (Sudo et al., 2019). Although the above studies demonstrate that miRNA has good sensitivity and specificity for diagnosing EC, the short length of miRNA fragments (only 19–24 nt) makes it difficult to distinguish from other similar sequences during detection, resulting in poorer specificity. In contrast, DNA methylation has a significant advantage over miRNA in terms of better specificity. Therefore, in future research on early diagnosis of EC, integrating the various advantages of miRNA and DNA methylation to develop a combined diagnostic kit might be a more promising direction.