Magnetically Controlled Capsule Endoscopy for Esophageal Varices: Systematic Review and Meta-Analysis

Abstract

Background: Magnetically controlled capsule endoscopy (MCCE) has shown promise in upper gastrointestinal evaluation and is a potentially less invasive alternative to esophagogastroduodenoscopy (EGD). We performed a systematic review and meta-analysis aiming to measure its diagnostic performance compared to EGD for esophageal varices. Methods: Our protocol was registered on PROSPERO (CRD420251081967). A systematic search of multiple databases was conducted through July 2025 for studies assessing the diagnostic performance of MCCE compared to EGD for EV. The primary outcomes were sensitivity and specificity. Secondary outcomes included the area under the curve (AUC), likelihood ratios, diagnostic odds ratio, and safety. Pooled effect estimates were calculated using a random effects model and expressed as proportions with 95% confidence intervals (CI). Heterogeneity was assessed using the I² statistic. Results: Five prospective studies with 795 patients (68.8% male, mean age of 55.1) were included. The pooled sensitivity and specificity were 87.1% (95% CI: 68.5–95.4) and 95.2% (95% CI: 88.4–98.1), respectively, with an AUC of 0.97. Following subgroup analysis of cirrhotic patients, pooled sensitivity and specificity were 96.0% (95% CI: 93.6–97.5; I² = 43%) and 95.2% (95% CI: 84.2–98.6; I² = 88.4%), respectively. Furthermore, MCCE use with detachable strings increased sensitivity to 96% (95% CI: 93.7–97.5; I² = 0%) and specificity to 96.3% (95% CI: 87.2–99.0; I² = 76.8%). Only four adverse events occurred across the five studies. Conclusions: Our meta-analysis demonstrated the high sensitivity, specificity, and diagnostic accuracy of MCCE, along with a favorable safety profile. Further large-scale trials are needed to validate our findings.

1. Introduction

Esophagogastroduodenoscopy (EGD), the primary modality for upper gastrointestinal (GI) tract evaluation, is often perceived as invasive and uncomfortable by patients. Capsule endoscopy (CE) offers a minimally invasive and patient-friendly alternative, particularly for small bowel disorders. However, its application in upper GI pathology is limited by rapid esophageal transit and uncontrolled passage through the stomach, resulting in incomplete visualization and reduced diagnostic accuracy [1,2,3].

Magnetically controlled capsule endoscopy (MCCE) overcomes these limitations by using an external magnetic field to precisely control capsule navigation and improve mucosal visualization, without the need for sedation or invasive instrumentation [4]. Since its introduction in 2006 [5], MCCE technology has advanced through multiple generations, incorporating various control mechanisms and capsule designs. Two main capsule types are used in MCCE platforms: the untethered MiroCam capsule, guided by external magnetic fields applied over the sternum, and detachable-string MCCE (ds-MCCE), where esophageal examination is performed while the capsule remains tethered, before transitioning to magnetic navigation in the stomach.

MCCE has proven to be a safe and feasible tool for detecting upper GI pathology [6]. Clinical studies comparing MCCE with EGD have shown comparable diagnostic accuracy in evaluating upper GI conditions [7,8,9,10,11]. A recent meta-analysis further demonstrated that MCCE is well tolerated and exhibits high sensitivity in detecting gastric lesions [12], supporting its clinical utility as a non-invasive diagnostic alternative for gastric disease assessment. MCCE is also increasingly being investigated for esophageal conditions, including esophageal varices (EV) [13,14].

Current guidelines recommend periodic surveillance for EV in patients with advanced chronic liver disease to reduce the risk of life-threatening variceal bleeding [15]. MCCE offers a minimally invasive and more comfortable option for evaluating EV, potentially improving patient adherence to surveillance. It may also help reduce healthcare system pressures by eliminating the need for sedation, recovery facilities, and specialized endoscopy teams, thus providing a more resource-efficient alternative to EGD. While a recent meta-analysis has evaluated the utility of MCCE in diagnosing esophageal conditions [16], data specifically addressing its role in screening and grading EV remain limited. Although early studies are promising, there remains a need to systematically assess the diagnostic performance of MCCE for EV compared to EGD. Therefore, we performed a systematic review and meta-analysis that aims to evaluate the diagnostic utility of MCCE for identifying EV.

2. Results

2.1. Search Results and Population Characteristics

The initial literature search yielded 677 citations across the four databases. Following exclusion of duplicate records, 432 titles and abstracts were screened, with 66 being chosen for full-text screening. Finally, a total of five studies were eligible for inclusion (Figure 1) [13,14,17,18,19]. All five studies were prospective cohorts in design, and three were multicenter, while two were single-center. The studies comprised a total of 792 undergoing MCCE with 795 undergoing EGD as the reference standard. The mean age of the participants was 55.1 years, and 68.8% were males (

Table 1. Baseline characteristics of included studies.

Study	Design	Location	Time	Population, n	Age, Mean	Males, n (%)	Population
Beg 2019 [13]	Prospective	UK—Single-center	2016–2017	50	60.42	34 (64%)	Healthy volunteers, patients with prior history of BE or EV
Chen 2019 [14]	Prospective	China—Single-center	2018	25	52.73	18 (72)	Healthy volunteers, patients with suspected esophageal disease
Jiang 2024 [17]	Prospective	China—Multicenter	2021–2022	582	56.51	398 (68.4)	Advanced chronic liver disease
Wang 2021 [18]	Prospective	China, UK— Multicenter	2018–2019	105	50.8	80 (76)	Advanced chronic liver disease
Ching 2019 [19]	Prospective	UK—Multicenter	2016–2017	33	NA *	25 (75.8)	Upper GI bleeding
Total				795	55.115	555 (68.8)

* Median age 60, IQR 24 years, BE: Barrett’s esophagus, EV: Esophageal varices.

). Investigator blinding was employed by all five studies.

2.2. Quality Assessment

In Beg et al. and Chen et al., a high risk of bias was assigned to the patient selection domain because a case–control study design was not avoided, as both of these patients enrolled healthy volunteers (controls) [13,14]. Applicability concerns were assigned to patient selection in Ching et al. due to the enrollment of patients with upper GI bleeding (Figure 2) [19].

2.3. Meta-Analysis Outcomes

2.3.1. Primary Outcome: Sensitivity and Specificity

Based on five studies with 792 patients [13,14,17,18,19], the pooled sensitivity and specificity using the bivariate model were 87.1% (95% CI: 68.5–95.4) and 95.2% (95% CI: 88.4–98.1), respectively. The correlation coefficient (ρ) was −0.26, indicating a weak inverse relationship such that studies with higher sensitivity tended to have slightly lower specificity, and heterogeneity measured by I² was 24.3% (

Table 2. Pooled sensitivity and specificity using the bivariate random effects model. Heterogeneity was low, based on the I²% of 24.3%.

	Estimate	Lower Bound	Upper Bound
Sensitivity	87.10%	68.50%	95.40%
Specificity	95.20%	88.40%	98.10%
Heterogeneity (I2)	24.30%

). Following univariate analysis, the pooled sensitivity and specificity were 90.4% (95% CI: 76.6–96.4) and 97.2% (95% CI: 91.6–99.1), respectively (Figure 3).

2.3.2. Subgroup Analysis

Two studies with 687 patients reported on the test performance in patients with liver cirrhosis. In that subgroup, the pooled sensitivity and specificity were 96.0% (95% CI: 93.6–97.5; I² = 43%) and 95.2% (95% CI: 84.2–98.6; I² = 88.4%), respectively (Figure 4).

Additionally, three studies with 712 patients reported on MCCE used with detachable strings. Subgroup analysis revealed a pooled sensitivity of 96% (95% CI: 93.7–97.5; I² = 0%) and a pooled specificity of 96.3% (95% CI: 87.2–99.0; I² = 76.8%) (Figure 5).

2.3.3. Secondary Outcomes

SROC curve and AUC: Using a bivariate random effects model, an SROC curve was constructed to evaluate the overall diagnostic accuracy of MCCE compared to EGD for the detection of EV (Figure 6). The summary point on the SROC curve corresponded to a pooled sensitivity of 87.1 and a pooled specificity of 95.2. The AUC was 0.97, while the partial area under the curve (pAUC) was 0.85.

Likelihood ratios: The pooled PLR for MCCE in detecting EV was 18.46 (95% CI: 7.12–47.91; I² = 62.7%), while the pooled NLR was 0.14 (95% CI: 0.05–0.37; I² = 86.6%) (Figure 7).

Diagnostic odds ratio: The pooled diagnostic odds ratio (DOR) for MCCE in the detection of EV was 177.40 (95% CI: 37.4–841.26, I² = 73.2%) (Figure 8).

Safety: Only four adverse events related to MCCE were reported by Jiang et al., while zero adverse events related to MCCE were reported in the remaining four studies. These were capsule retention (n = 2) as well as syncope (n = 1) and hemorrhoidal bleeding (n = 1) related to gastrointestinal preparation.

3. Validation of Meta-Analysis

3.1. Sensitivity Analysis

Sensitivity analysis did not reveal a significant difference in the pooled sensitivity or specificity with individual study exclusion, indicating robustness of the pooled estimates (Figure 9).

3.2. Heterogeneity

We assessed the dispersion of the calculated rates using the confidence interval (CI) and I² percentage values. Based on the 95% CI and I²% values across included studies, the overall effects distribution was low with regard to the primary outcome of sensitivity and specificity, but substantial heterogeneity was noted across the subgroup analyses and secondary outcomes.

4. Discussion

Our meta-analysis evaluating the diagnostic performance of MCCE for EV detection demonstrated excellent test performance and tolerability. The high sensitivity, specificity, and AUC support the potential utility of MCCE for EV screening and surveillance. Subgroup analyses further validate these findings, with sensitivity and specificity exceeding 95% in cirrhotic populations and in studies utilizing ds-MCCE. The favorable safety profile supports its role as a reliable, minimally invasive alternative to EGD.

Although CE has been explored as a minimally invasive test for screening EV, current evidence suggests it has limited diagnostic accuracy. Several studies have reported moderate diagnostic performance, with sensitivity ranging from 76% to 85.8% and specificity from 80.5% to 93% [1,20,21]. As a result, it is not a recommended screening modality as per current guidelines and systematic reviews [15,20]. MCCE addresses key limitations of conventional CE by enabling controlled and repeatable visualization of the esophageal mucosa. Unlike traditional capsules that passively transit the esophagus, which often results in incomplete views and reduced sensitivity, MCCE allows precise positioning and enhances diagnostic yield and reliability [3,22]. This is demonstrated by results from a recent meta-analysis by Yang et al., which demonstrated high diagnostic accuracy of MCCE for esophageal diseases, with a pooled sensitivity of 95%, specificity of 97%, and an AUC of 0.99 [16]. In comparison, our study reported a slightly lower sensitivity of 87.1% and specificity of 95.2%, likely due to differences in sample size, the number of studies included, and the focus on EV rather than general esophageal pathology, as small EV often present as subtle, flat, or minimally elevated submucosal lesions and can be more difficult to visualize with MCCE. Additionally, the absence of insufflation may limit gastric and esophageal distension, reducing exposure of submucosal contours and making their detection more challenging than for superficial mucosal abnormalities [23].

The addition of a tethered capsule in ds-MCCE further improves maneuverability, enabling fine-tuned steering, rotation, and prolonged dwell time in anatomically narrow or mobile regions, such as the esophagus and cardia. This enhanced operator control has been shown to result in superior mucosal visualization and increased detection of esophageal lesions compared to conventional MCCE [24,25]. Our subgroup analysis aligned with these findings, demonstrating higher sensitivity and specificity in the ds-MCCE group compared to conventional MCCE. Additional evidence supports the diagnostic utility of advanced capsule endoscopy systems. A prospective, blinded study by Yang et al. evaluating ds-MCCE also demonstrated high sensitivity and specificity, with patients describing the procedure as both comfortable and convenient [26]. Similarly, Jiang et al. found that ds-MCCE successfully visualized the upper and mid GI tract in 95.45% of patients and exhibited a high diagnostic yield, with a per-lesion sensitivity of 96.2% for upper GI diseases, with all procedures being well tolerated and associated with low discomfort scores [27]. Furthermore, a multicenter trial revealed that the diagnostic performance of cable-transmission MCCE, which utilizes a linked electric wire, is comparable to that of EGD in both the completeness of upper GI tract examinations and lesion detection [28]. These studies highlight the potential for even greater diagnostic yield with future technological refinements in capsule design and maneuverability.

Beyond diagnostic performance, several of the included studies reported important secondary outcomes. Jiang et al. showed that ds-MCCE was accurate not only for detecting EV but also for identifying high-risk varices and gastric varices, with accuracy approaching 97% [17]. Furthermore, studies by Wang et al. and Chen et al. highlighted higher patient satisfaction, minimal discomfort, and absence of adverse events during ds-MCCE procedures [14,18]. Beg et al. also found MACE to be significantly more comfortable than EGD, reporting higher preference and tolerance scores [13]. Additionally, Ching et al. evaluated MCCE in the acute setting of upper GI bleeding and demonstrated that the modality not only had a high diagnostic yield for focal lesions but also correctly identified patients safe for discharge, suggesting a valuable role in emergency triage. Notably, MCCE was also able to identify small bowel sources of bleeding in 18% of patients, an added diagnostic benefit not achievable with EGD alone [19]. In terms of procedural metrics, visualization quality was generally high across studies. For example, Beg et al. reported that the capsule could be held in the esophagus for a mean of 190 s (up to 634 s) using external magnetic control, enabling detailed mucosal inspection [13]. Similarly, Ching et al. found that MCCE achieved excellent views of the gastric body and antrum in most cases, though visibility was somewhat reduced in the esophagus and duodenal bulb [19]. In the study by Chen et al., image quality was graded between 8 and 10 out of 10, with clear visualization of key landmarks and minimal obscuration by bubbles or mucus [14]. These patient-centered and technical outcomes underscore MCCE’s promise as a feasible outpatient or community-based tool that may help expand access to variceal screening, reduce burden on endoscopy services, and improve adherence to surveillance protocols. On the other hand, MCCE can be more technically complex and time-consuming. This was demonstrated by Jiang et al., where the median examination time for the esophagus and stomach was 4.74 min and 15.78 min, respectively. In comparison, the examination time for esophagus, stomach, and duodenum using EGD was only 5.50 min [17]. Furthermore, the inability to perform biopsies or therapeutic interventions remains a major limitation of MCCE. Although rare, another limitation of MCCE is the inability to swallow the capsule for some patients. This was reported in three patients (6%) by Beg et al. and two (1.8%) by Wang et al. [13,18]. Nonetheless, ongoing technological advancements, as demonstrated by ds-MCCE, hold promise for overcoming current procedural constraints and further enhancing the diagnostic role of capsule-based endoscopy in clinical practice.

A major strength of our analysis is that, to our knowledge, it represents the first systematic review and meta-analysis specifically evaluating the performance of MCCE in the diagnosis of EV. It addresses a key literature gap and provides important insights into potential non-invasive alternatives to EGD. Methodologically, the study adhered strictly to PRISMA 2020 guidelines and employed a comprehensive search strategy across multiple databases. Robust statistical methods were applied to ensure the validity and reliability of pooled estimates. Notably, all included studies were prospective, blinded, and exhibited a low overall risk of bias, with the majority being multicenter in design, enhancing the generalizability of the findings. Sensitivity analyses demonstrated that the exclusion of individual studies did not significantly alter the results, supporting their robustness. Moreover, the primary outcomes exhibited low heterogeneity, indicating minimal between-study variability. Furthermore, subgroup analysis based on study population and capsule design was performed to offset study heterogeneity.

The study also has several limitations, with many being inherent to most meta-analyses. First, only five studies met the eligibility criteria, encompassing a total of 795 patients, which limited the statistical power for meta-regression and subgroup analyses, and precluded formal assessment of publication bias. Moreover, the small number of included studies restricted our ability to perform bivariate random effects analyses within subgroups, which may affect the robustness and generalizability of the subgroup findings. Second, all included studies were conducted in China and the United Kingdom, potentially limiting the geographic generalizability of the findings. Third, the study population was predominantly male (68.8%), which may reduce the applicability of the results to female patients. Fourth, two studies exhibited potential selection bias, and one raised concerns regarding applicability. Consequently, unidentified sources of heterogeneity may exist but could not be explored due to limitations in the available data. Fifth, variability in procedural techniques and in the time interval between MCCE and EGD likely contributed to heterogeneity across studies. Finally, the inconsistent reporting of important outcomes such as visualization and tolerance and lack of cost-effectiveness analyses precluded quantitative synthesis of these parameters. Future research should incorporate standardized metrics for comfort, satisfaction, transit control, visualization scoring, and cost-effectiveness to enable robust comparisons across technologies and settings.

Nevertheless, our meta-analysis highlights the growing potential of MCCE as a non-invasive alternative to EGD for the diagnosis and surveillance of EV, with greater patient tolerability and satisfaction and reduced discomfort. This potential is further enhanced by recent technological advancements, such as the introduction of ds-MCCE. While the findings are promising, their clinical applicability is limited by the small number of included studies and variations in study protocols. Future large-scale, multicenter trials with cost-effectiveness analysis are warranted to validate these results across diverse populations and healthcare settings.

5. Methods

5.1. Protocol and Registration

This meta-analysis adheres to the guidelines of the updated preferred reporting items for systematic reviews and meta-analyses (PRISMA) 2020 statement [29]. The review protocol was prospectively registered on PROSPERO (CRD420251081967).

5.2. Search Strategy and Study Selection

A systematic search was conducted in PubMed, Web of Science, EMBASE, and Scopus up to 1st July 2025, to identify relevant studies evaluating the diagnostic performance of MCCE compared to EGD for EV. Full texts of potentially eligible studies were retrieved and assessed for inclusion. The inclusion criteria included prospective and retrospective interventional and observational studies in the form of full-text publications that reported diagnostic performance data with EGD as the reference standard. Meanwhile, the exclusion criteria included review articles, case reports, case series with fewer than five patients, conference proceedings, and missing diagnostic performance data. To ensure a comprehensive and inclusive search strategy, we used a combination of medical subject heading terms and free-text keywords related to MCCE and EV. No date restrictions were used (full search strategy in Supplementary Materials). Two investigators, T.A and P.M., independently screened citations, resolving discrepancies with a third investigator J.G.

5.3. Data Extraction and Outcome Measures

Data were extracted into a Google Sheets spreadsheet by two independent investigators (T.A. and P.M.), with a third investigator (J.G.) resolving discrepancies. We collected data from each eligible study, including author’s name, publication year, country, type of publication, study design, number of patients, demographics, test methodology, and outcome. The primary outcomes of interest were the sensitivity and specificity of MCCE. Secondary outcomes were the area under the summary receiver operating characteristic (SROC) curve (AUC), positive and negative likelihood ratios (PLR and NLR), and adverse events related to MCCE. Sensitivity analysis was performed by assessing the impact of individual study exclusion on the pooled estimate. Subgroup analysis was performed for cirrhosis patients and ds-MCCE.

5.4. Quality Assessment

The Quality Assessment of Diagnostic Accuracy Studies, Version 2 (QUADAS-2) was used to assess study quality and bias [30]. Two independent investigators (T.A. and P.M.) conducted the assessments, with a third investigator (J.G.) resolving disagreements.

5.5. Statistical Analysis

The meta-analysis was conducted using the mada and metafor packages on R (version 4.5.0). Test performance data from individual studies were pooled to estimate effect sizes using the bivariate and univariate random effects models. Pooled values were calculated and presented as proportions with corresponding 95% confidence intervals (CIs). Forest plots were used to visualize the pooled estimates. Using the bivariate random effects model, the correlation coefficient (ρ) was also calculated, estimating how sensitivity and specificity change in relation to each other across studies, reflecting any trade-off in test performance. Heterogeneity was assessed using the I² statistic following Cochrane Handbook guidelines: 0–40% may indicate low heterogeneity, 30–60% moderate, 50–90% substantial, and 75–100% considerable heterogeneity [31]. A sensitivity analysis was performed to assess the impact of individual studies on the cumulative effect size for the primary outcome. Ages reported by the authors as median and interquartile range (or ranges) were converted into means and standard deviations [32]. The mean age was calculated as a mean of means. A p-value of 0.05 or lower was considered as statistically significant unless otherwise stated. A formal assessment of publication bias or certainty in the body of evidence was not performed as fewer than 10 studies were available.