Dysplasia-Stratified Surveillance Identifies Optimal Strategies for Preventing Esophageal Adenocarcinoma in Barrett’s Esophagus: An Incidence-Based Cost-Effectiveness Model
Department of Occupational Health, Kitasato University Graduate School of Medical Sciences, Kanagawa, Japan
Advanced Research Promotion Center, Health Sciences University of Hokkaido, Hokkaido, Japan
*E-mail: kowadaa@gmail.com; kowada.akiko@kitasato-u.ac.jp;Abstract
Background
The risk of esophageal adenocarcinoma (EAC) in Barrett’s esophagus (BE) varies substantially by segment length and dysplasia grade, yet optimal surveillance strategies remain uncertain.
Aims
To evaluate the cost–effectiveness and health impacts of dysplasia–stratified EAC surveillance strategies for the Japanese BE population.
Methods
We developed a state–transition model comparing endoscopy, sponge test, breath test, and miRNA test with no surveillance from a healthcare payer perspective. Non–invasive strategies were evaluated as primary surveillance tools, with positive results triggering confirmatory endoscopy. A scenario analysis assessed AI–assisted endoscopy. Five BE populations of 50–year–old individuals were modeled over a lifetime: ultra–short segment BE (USSBE); short-segment BE (SSBE); long–segment nondysplastic BE (LSBE–NDBE); LSBE with low–grade dysplasia (LSBE–LGD); and LSBE with high–grade dysplasia (LSBE–HGD). Each modality was evaluated at surveillance intervals of 1, 2, 3, 4, 5, or 10 years. Primary outcomes included net monetary benefits, costs, quality–adjusted life–years, incremental cost–effectiveness ratios, and EAC deaths. Deterministic and probabilistic sensitivity analyses assessed parameter uncertainty.
Results
Surveillance was not cost–effective for USSBE, SSBE, or LSBE–NDBE. For LSBE–LGD, annual endoscopy was most cost–effective, averting 83 EAC deaths per 10,000 individuals over a lifetime. For LSBE–HGD, annual breath test was most cost–effective, averting 295 deaths. Cost–effectiveness was driven primarily by subgroup–specific EAC incidence and test adherence.
Conclusions
Annual endoscopy is cost–effective for LGD, and annual breath test is cost–effective for HGD, whereas surveillance offers limited benefit for NDBE. These findings support a dysplasia–specific framework for surveillance in LSBE and highlight the limited value of routine surveillance for NDBE.
Article notes
Competing Interest Statement
The authors have declared no competing interest.
Funding Statement
This study did not receive any funding.
1.Introduction
Barrett’s esophagus (BE) is the principal precursor to esophageal adenocarcinoma (EAC), a malignancy with rising incidence and poor prognosis when diagnosed at advanced stages. Globally, EAC accounted for approximately 85,700 new cases in 2020, representing about 14% of all esophageal cancers, with the highest incidence observed in high–income regions such as North America, Western and Northern Europe, and Australia [1]. Notably, EAC began increasing sharply in Europe and the United States in the 1960s, marking the beginning of a rapid epidemiological transition [2]. In contrast, EAC remains relatively uncommon across Asia; however, the prevalence of BE has increased in several East Asian countries over the past two decades. In Japan, although squamous cell carcinoma remains predominant, the incidence of EAC has shown a gradual upward trend, suggesting that Japan may be entering a phase similar to that experienced by Western countries several decades earlier [3].
In Western countries, major gastroenterological societies, including the American College of Gastroenterology (ACG) [4], the American Society for Gastrointestinal Endoscopy (ASGE) [5], and the British Society of Gastroenterology (BSG) [6], recommend structured endoscopic surveillance for patients with BE. Surveillance intervals are primarily determined by dysplasia grade, with segment length used to further refine intervals in nondysplastic BE (NDBE). In parallel with these guideline–based strategies, several emerging technologies have been developed to enhance early detection of dysplasia and EAC. These innovations serve two complementary roles: improving the diagnostic performance of endoscopy, exemplified by artificial Intelligence (AI)–assisted systems capable of real–time identification of subtle neoplastic changes [7]; and reducing the burden of endoscopic examinations through minimally invasive pre–endoscopic triage tools such as sponge–based cell–collection devices (e.g., Cytosponge) [8], breath–analysis platforms that detect volatile organic compounds [9], and circulating biomarker assays including miRNA–based tests [10]. Together, these technologies aim to increase detection sensitivity while improving accessibility and patient acceptability of BE surveillance.
In Japan, however, a standardized national surveillance framework for BE has not been established. Although BE is frequently detected partly due to the broader endoscopic definition that classifies any visible columnar epithelium as BE, surveillance practices vary widely among institutions, and most cases represent ultrashort–segment BE (USSBE) [3]. Recent epidemiological analyses suggest that the incidence of EAC in Japan is projected to increase in the coming decades [3]. Japan has long maintained a nationwide endoscopic screening program for gastric cancer, providing regular opportunities for incidental detection of esophageal neoplasia among individuals over 50 years of age. As a result, early detection of EAC has relied largely on opportunistic esophagogastroduodenoscopy rather than structured surveillance. However, as gastric cancer incidence continues to decline due to decreasing Helicobacter pylori infection rates and the widespread implementation of eradication therapy [11], accompanied by a marked increase in the prevalence of reflux esophagitis, the current screening system is expected to be simplified [3]. This may reduce opportunities for incidental EAC detection, even if gastric cancer surveillance is maintained after H. pylori eradication. Consequently, establishing a dedicated and evidence–based surveillance strategy for BE has become increasingly important.
Given the rising prevalence of BE, the anticipated increase in EAC incidence, and the absence of a unified surveillance strategy in Japan, the cost–effectiveness of stratified EAC surveillance approaches warrants rigorous evaluation to inform evidence–based early detection strategies tailored to the Japanese BE population. Therefore, the aim of this study is to evaluate the cost–effectiveness of optimal stratified EAC surveillance strategies for the BE population in Japan, with surveillance intervals defined by segment length and dysplasia grade.
2.Methods
2.1Model Overview
We developed a state–transition model to evaluate the cost–effectiveness of stratified surveillance strategies for EAC among individuals with BE in Japan. Five BE subpopulations were modeled independently based on segment length and dysplasia grade: ultra–short segment BE (USSBE), short–segment BE (SSBE), long–segment BE (LSBE) without dysplasia (LSBE-NDBE), LSBE with low–grade dysplasia (LSBE-LGD), and LSBE with high–grade dysplasia (LSBE-HGD) (Figure 1). Each subgroup was simulated separately to reflect its distinct annual incidence of EAC [12,13].
The model followed a hypothetical cohort of 50–year–old individuals with BE over a lifetime from a healthcare payer perspective. We selected age 50 because it approximates the median age at BE diagnosis and allows evaluation of lifetime surveillance benefits. Surveillance strategies included conventional endoscopy, sponge test, breath test, and miRNA test, each evaluated at intervals of 1, 2, 3, 4, 5, or 10 years. Non–invasive strategies were modeled as primary surveillance tools followed by confirmatory endoscopy. A no–surveillance strategy served as the reference comparator. A one–year cycle length with half–cycle correction was applied, and both costs and quality–adjusted life–years (QALYs) were discounted at 3% annually.
2.2Model Structure
Five independent Markov models were constructed, one for each BE subgroup, to reflect subgroup–specific EAC incidence without assuming unobserved natural–history transitions. Individuals entered the model in their corresponding BE health state (NDBE, LGD, or HGD), underwent surveillance according to the assigned strategy, and remained in that state unless they developed EAC or died.
No transitions between BE states were assumed because the model relied solely on observed EAC incidence and mortality rather than hypothetical natural–history progression pathways. No preclinical disease state or interval cancer construct was included; all cancers were assumed to occur and be diagnosed within the same annual cycle. Mortality included stage–specific EAC survival [14] and background mortality from Japanese life tables [15]. Primary outcomes included net monetary benefits (NMBs), costs, QALYs, incremental cost–effectiveness ratios (ICERs), and EAC deaths. A schematic representation of health states and allowable transitions is shown in Figure 2.
2.3Surveillance Strategies
Four surveillance modalities were evaluated: conventional endoscopy, sponge test, breath test, and miRNA test. Each modality was assessed at surveillance intervals of 1, 2, 3, 4, 5, or 10 years, with no surveillance as the reference. AI–assisted endoscopy was evaluated in scenario analyses as an enhanced form of conventional endoscopy.
Conventional endoscopy represented the current standard of care in Japan. Sponge test, breath test, and miRNA test were modeled as non–invasive primary surveillance tools. A positive non–endoscopic test triggered confirmatory endoscopy and subsequent management according to stage–specific treatment pathways in the Japanese Guidelines for the Diagnosis and Treatment of Esophageal Cancer [16]. Individuals with negative results continued routine surveillance at the assigned interval.
2.4Diagnostic Performance
Sensitivity and specificity for conventional endoscopy were obtained from meta–analyses and large observational cohorts [17]. Sensitivity and specificity for sponge test, breath test, miRNA test, and AI–assisted endoscopy were extracted from published studies [8,9,10,18].
False–negative results were modeled as cancers diagnosed clinically within the same cycle, with stage distribution based on clinically detected EAC. Adherence rates for conventional endoscopy, AI–assisted endoscopy, and breath test were obtained from published sources [19], whereas adherence for sponge test and miRNA test was assumed. All adherence parameters were varied widely (10–90%) in sensitivity analyses.
2.5Model Inputs
Model inputs included subgroup–specific annual EAC incidence [12,13], diagnostic performance parameters, costs, health–state utilities, and mortality rates. Annual EAC incidence was assigned separately for each BE subgroup (USSBE, SSBE, LSBE–NDBE, LSBE–LGD, LSBE–HGD). Sensitivity, specificity, and adherence were assigned independently for each surveillance modality [8,9,10,17,18]. Costs reflected testing and cancer treatment expenditures [20]. Health–state utilities were derived from published literature [21,22]. Stage–specific EAC survival was based on Japanese data [14], and background mortality was derived from Japanese life tables [15]. Costs and QALYs were discounted at 3% annually [23].
2.5.1Costs
Costs were derived from the Japanese national fee schedule [20] and converted to U.S. dollars using the 2024 OECD purchasing power parity exchange rate (95.1 yen per US$1) [24]. Costs for sponge test, breath test, AI–assisted endoscopy, and miRNA test were not available in Japan and were therefore treated as assumptions, with ranges explored in sensitivity analyses.
2.5.2Health–State Utilities
Utilities were assigned to mutually exclusive health states representing the clinical course of BE and EAC, including NDBE, LGD, HGD, four stage–specific EAC states (I–IV), and three post–treatment states [21,22]. Death was modeled as an absorbing state. Utility values were derived from published literature and discounted at 3% [23].
2.6Base–Case Analysis
The base–case analysis estimated costs, QALYs, ICERs, and NMBs. Strategies were ordered by increasing QALYs, and extended dominance was applied when appropriate. A willingness–to–pay threshold of US$50,000 per QALY gained was used to determine cost–effectiveness [25].
2.7Markov Cohort Analyses
Using a Markov cohort model, we estimated, for each surveillance strategy, the cumulative 10–year and lifetime numbers of EAC deaths averted and the cumulative 10–year and lifetime additional costs per EAC death averted, compared with no surveillance. Analyses were conducted separately for hypothetical cohorts of 10,000 LSBE–LGD patients and 10,000 LSBE–HGD patients aged 50 years.
For each cohort, EAC deaths averted per 10,000 patients were calculated by multiplying the difference in the cumulative probability of EAC death between each surveillance strategy and no surveillance by 10,000. Additional costs per EAC death averted were calculated by multiplying the difference in total discounted costs between each surveillance strategy and no surveillance by 10,000 and dividing this value by the number of EAC deaths averted in the corresponding 10,000–patient cohort.
2.8Scenario Analyses
Scenario analyses evaluated AI–assisted endoscopy by substituting its diagnostic performance parameters for those of conventional endoscopy within each model. This approach allowed us to explore how improvements in sensitivity and specificity attributable to AI would influence overall cost–effectiveness across the surveillance strategies.
2.9Sensitivity Analyses
One–way sensitivity analyses varied key clinical and economic parameters across plausible ranges, including test sensitivity and specificity, surveillance and treatment costs, health–state utilities, and subgroup–specific EAC incidence. We also varied the starting age of surveillance from 40 to 80 years to evaluate how age at initiation influenced cost–effectiveness outcomes. Deterministic results were summarized using tornado diagrams.
A probabilistic sensitivity analysis was conducted using 10,000 Monte Carlo simulations. Parameter uncertainty was represented using beta distributions for test performance, subgroup–specific EAC incidence, and utilities, and gamma distributions for costs. For each simulation, lifetime costs and QALYs were recalculated, and cost–effectiveness acceptability curves were generated at a willingness–to–pay (WTP) threshold of US$50,000 per QALY gained.
All analyses were conducted using TreeAge Pro 2026 (TreeAge Software, Williamstown, MA). This economic evaluation followed the CHEERS 2022 reporting guidelines [26].
2.10Model Validation
Model validation included face, internal, and external validation. Face validation assessed whether the model structure, assumptions, and parameter inputs were consistent with established frameworks for BE and EAC modeling. Internal validation evaluated whether simulated transitions, incidence patterns, and long–term outcomes behaved as expected across all five BE subgroups. External validation compared model–generated estimates of EAC incidence, stage distribution, and survival with published epidemiologic data. Together, these validation checks supported the appropriateness and credibility of the model for evaluating stratified surveillance strategies.
3.Results
3.1Base–case results
Base–case outcomes differed markedly across BE subgroups, driven primarily by variation in annual EAC incidence.
USSBE
No surveillance was the preferred strategy, yielding the highest NMB (US$869,600) (Table S1). Annual endoscopy increased costs substantially (US$3,942 vs US$28) with only negligible QALY gains, lowering NMB to US$865,697. Annual breath test and annual miRNA test produced even lower NMBs.
SSBE
No surveillance again produced the highest NMB (US$950,019) (Table S2). Annual endoscopy increased costs (US$4,117 vs US$225) with minimal QALY improvement, reducing NMB to US$946,232. Across both short–segment groups, surveillance strategies increased costs with only marginal benefits and were not cost–effective at a WTP of US$50,000 per QALY.
LSBE–NDBE
No surveillance remained the preferred strategy (NMB US$938,821) (Table S3, Figure 3A). Annual endoscopy raised costs (US$5,484 vs US$1,770) and modestly improved QALYs, but resulted in a lower NMB (US$935,939). Annual breath test and annual miRNA test yielded even lower NMBs. None of the surveillance strategies were cost–effective.
LSBE–LGD
Annual endoscopy was the most cost–effective strategy, producing the highest NMB (US$841,993) (Table S4, Figure 3B). Although no surveillance had lower costs, it also produced fewer QALYs, resulting in a slightly lower NMB. Annual breath test and annual miRNA test generated marginally higher QALYs but substantially lower NMBs, indicating that their additional benefits did not justify their higher costs.
LSBE–HGD
Annual breath test was the preferred strategy, with the highest NMB (US$598,066) and an ICER of US$16,318 per QALY gained (Table S5, Figure 3C). Annual miRNA test produced the highest QALYs but was not cost–effective due to its high ICER and lower NMB.
Overall
Surveillance was not cost–effective in low–risk subgroups (USSBE, SSBE, LSBE–NDBE). In contrast, annual endoscopy for LSBE–LGD and annual breath test for LSBE–HGD offered the greatest value in higher–risk populations.
3.2EAC deaths averted and additional costs per death averted
EAC deaths averted and additional costs per EAC death averted were evaluated for 10,000 LSBE–LGD and 10,000 LSBE–HGD patients over 10 years and a lifetime.
In LSBE–LGD, annual endoscopy prevented 24 EAC deaths within 10 years and 83 EAC deaths over a lifetime (Table 2, Figure S6), with lifetime additional costs of US$349,308 per EAC death averted (Table 3, Figure S8). Annual breath test and annual miRNA test prevented slightly more EAC deaths (92 and 97 lifetime deaths, respectively) but required substantially higher additional costs per EAC death averted.
In LSBE–HGD, annual breath test prevented 209 EAC deaths within 10 years and 295 deaths over a lifetime (Table 2, Figure S7), costing US$5,006 over 10 years but ultimately saving US$1,394 over a lifetime per EAC death averted (Table 3, Figure S9). Although annual miRNA test prevented the most deaths (311 over a lifetime), it did so at substantially higher additional costs per EAC death averted. These findings indicate that the absolute mortality benefit of surveillance rises sharply with baseline EAC incidence, with annual breath test offering the most favorable balance of EAC death reduction and cost in HGD.
3.3Scenario analyses
When AI–assisted endoscopy was evaluated as an alternative to conventional endoscopy, no surveillance remained cost–effective in USSBE, SSBE, and LSBE–NDBE at a WTP threshold of US$50,000 per QALY (Table S5-S8).
In LSBE–LGD, annual AI–assisted endoscopy yielded the highest NMB (US$842,747), whereas in LSBE–HGD, annual breath test following AI–assisted endoscopy produced the highest NMB (US$603,334) (Table S9 and Table S10). These effects were driven by the higher sensitivity and specificity of AI–assisted endoscopy, which increased early detection while reducing false positives in higher–risk subgroups.
3.4Sensitivity analyses
One–way sensitivity analyses indicated that cost–effectiveness was primarily driven by subgroup–specific EAC incidence and test adherence.
In LSBE–LGD, annual endoscopy became cost–effective relative to no surveillance when annual EAC incidence exceeded 0.009 (Figure S1).
In LSBE–HGD, annual breath test was favored over annual endoscopy when endoscopy adherence exceeded 83.5% and breath–test adherence remained below 88.8% (Figure S2). Annual miRNA test became more cost–effective than annual breath test when breath–test adherence fell below 83.7% (Figure S3).
Cost–effectiveness acceptability curves showed a 79% probability that no surveillance is cost–effective in LSBE–NDBE (Figure 4A), a 36% probability that annual endoscopy is cost–effective in LSBE–LGD (Figure 4B and Figure S4), and a 29% probability that annual breath test is cost–effective in LSBE–HGD (Figure 4C and Figure S5) at a WTP threshold of US$50,000 per QALY gained.
4.Discussion
4.1Principal Findings
This study demonstrates that stratified surveillance strategies for BE, defined by segment length and dysplasia grade, provide the most cost–effective approach for the Japanese population. By incorporating Japan–specific incidence data for USSBE and SSBE and applying Western subgroup–specific incidence estimates for LSBE (NDBE, LGD, and HGD) derived from high–quality meta–analyses [12], we developed a transparent incidence–based framework across five clinically distinct subgroups. Using this approach, we identified that annual endoscopy is the optimal strategy for patients with LGD, whereas annual breath test is the most cost–effective strategy for those with HGD.
Across subgroups, differences in EAC deaths averted closely mirrored differences in baseline EAC incidence, indicating that incidence primarily determines the preferred surveillance strategy. As incidence increased, the optimal strategy shifted in a clear stepwise pattern: surveillance offered limited value in USSBE, SSBE, and LSBE–NDBE; annual endoscopy became optimal in LSBE–LGD; and annual breath testing became optimal in LSBE–HGD. This pattern reflects the strong dependence of surveillance benefit on underlying cancer risk and the diagnostic performance of available tests.
Because AI–assisted endoscopy improves sensitivity and specificity from 75% and 70% to 93% and 78%, respectively, it directly replaces standard endoscopy in the incidence range where endoscopic surveillance was previously optimal. This shift illustrates how improvements in test accuracy move the incidence thresholds that define the preferred strategy, further reinforcing the transparency and adaptability of an incidence–based modeling approach.
To our knowledge, this study represents the first fully integrated, risk–stratified cost–effectiveness analysis of EAC surveillance among patients with Barrett’s esophagus worldwide. Importantly, because LSBE incidence inputs are derived from Western literature, the model is not limited to Japan; substituting region–specific cost parameters would allow direct application to Western healthcare settings.
Japan provides a unique context for evaluating surveillance strategies because it has one of the world’s highest proportions of ultrashort–segment BE [3] and a lower baseline incidence of EAC compared with Western countries. Historically, early detection of EAC in Japan has relied on incidental findings during gastric cancer screening, routine health checkups, and other clinical encounters. As opportunities for incident detection are expected to decline [3], risk–stratified surveillance becomes increasingly important for maintaining early–stage diagnosis while avoiding unnecessary procedures in low–risk groups. Emerging technologies such as breath test and AI–assisted endoscopy may further enhance surveillance efficiency, particularly in intermediate– and high–risk groups. Overall, this study provides a policy–relevant foundation for developing surveillance recommendations tailored to Japan while offering a generalizable framework applicable to Western populations.
4.2Comparison with Previous Studies
Two cost–effectiveness analyses have previously evaluated BE surveillance. Kastelein et al. demonstrated that surveillance with endoscopic mucosal resection and radiofrequency ablation for HGD or early EAC, and oesophagectomy for advanced EAC, is cost–effective every 5 years for NDBE and every 3 years for LGD, based on a Dutch healthcare perspective and a willingness–to–pay threshold of €35,000 per QALY [27]. Vissapragada et al. further identified risk–stratified surveillance strategies as the most cost–effective, showing that biennial surveillance for long–segment BE (>2 cm) and 12–month surveillance for LGD provided the best value, and confirming the utility of dysplasia–based risk stratification in Australia [28]. These cost–effectiveness models simulated the natural history of Barrett’s esophagus to EAC and required numerous structural assumptions, including progression rates, dwell times, and transition probabilities, that vary widely across studies and contribute to inconsistent estimates of surveillance benefit, particularly in NDBE. Our findings additionally identified annual breath test for LSBE–HGD and annual endoscopy for LSBE–LGD as the most cost–effective strategies, and showed that AI–assisted endoscopy can serve as a cost–effective alternative to conventional endoscopy. Our study also employs an incidence–based framework anchored to empirically observed annual EAC incidence, avoiding assumptions about preclinical disease dynamics and providing a more transparent representation of cancer risk across clinically distinct subgroups.
Evaluating the cost–effectiveness of surveillance for USSBE and SSBE is essential because these segments represent the majority of BE cases in Japan [3]. This study is the first to demonstrate limited cost–effectiveness of surveillance for NDBE, including USSBE and SSBE, when empirical incidence is applied without additional structural assumptions. Our model also incorporates Western dysplasia–specific incidence estimates for LSBE [13], reflecting the limited availability of Japanese data and anticipating future increases in LGD and HGD, thereby supporting applicability to Western populations as well. Furthermore, our evaluation of emerging technologies, such as breath test and AI–assisted endoscopy, extends previous work by demonstrating their potential value in Japan and in higher–incidence settings [9,18]. Non–endoscopic examinations may further alleviate the surveillance burden for individuals with BE, particularly in settings where frequent endoscopy is difficult to sustain.
4.3Strengths and Limitations
This study has several methodological strengths. First, this study employs an incidence–based framework that does not rely on unobservable progression states or assumptions about transitions between nondysplastic BE, dysplasia, and preclinical cancer. By grounding the model in empirically observed annual EAC incidence for each subgroup, the analysis avoids structural uncertainty inherent in traditional multistate models and provides a transparent representation of cancer risk. Second, evaluating five clinically distinct BE subgroups independently allows surveillance strategies to be compared within risk strata, clarifying why surveillance yields limited benefit in low–risk groups and substantial benefit in high–risk groups. Third, the model incorporates modality–specific test performance and applies consistent probability distributions across parameters, enhancing transparency and reducing reliance on unverifiable assumptions. Finally, using a fixed cohort of 10,000 individuals per subgroup ensures comparability across analyses without requiring uncertain estimates of subgroup sizes in Japan. These features collectively clarify the stronger efficiency of surveillance for individuals with LSBE.
This study also has limitations. First, dysplasia–stratified EAC incidence estimates are not available in Japan. Second, reliable data on the population size of each BE subgroup in Japan are lacking [3], preventing estimation of national–level impacts. Third, several evaluated modalities, such as sponge test, breath test, miRNA test, and AI–assisted endoscopy, are not yet reimbursed under the Japanese national insurance system, requiring cost assumptions that may differ from future reimbursement prices. Fourth, test performance parameters were derived from heterogeneous Western studies [8,9,10,17,18], and real–world performance in Japan may differ. Finally, patient burden associated with undergoing surveillance tests was not incorporated into the model, although it is an important consideration, particularly for non–invasive modalities.
4.4Clinical and Policy Implications
The findings of this study have important implications for clinical practice and health policy. First, the clear stratification of surveillance benefit across Barrett’s esophagus subgroups supports a shift toward risk–based surveillance rather than uniform intervals. The limited value of surveillance in USSBE, SSBE, and LSBE–NDBE suggests that routine endoscopic surveillance may not be justified for these low–risk groups in Japan. Conversely, the strong cost–effectiveness observed in LSBE–LGD and LSBE–HGD reinforces the need for timely and intensive surveillance in higher–risk patients, consistent with their substantially elevated incidence of EAC.
Second, the results highlight the potential role of emerging non–endoscopic technologies, particularly breath test, which demonstrated favorable cost–effectiveness in high–risk groups. These modalities may reduce the burden of repeated endoscopy, improve patient acceptability, and expand access to surveillance in settings with limited endoscopic capacity. AI–assisted endoscopy may further enhance diagnostic accuracy and efficiency as evidence accumulates [7]. AI–assisted endoscopy, which surpasses the diagnostic performance of conventional endoscopy and offers greater cost–effectiveness, is likely to replace standard endoscopy in the incidence range where it is currently preferred, further improving the efficiency of surveillance programs.
Third, future integration of individualized risk–prediction tools could further enhance precision. Although segment length and dysplasia grade remain the strongest determinants of cancer risk, additional factors, such as age, sex, obesity, and GERD severity [29,30], could refine risk stratification and guide personalized surveillance intensity.
Fourth, the incidence–based and hybrid structure of the model, which uses Japan–specific incidence for USSBE and SSBE and Western subgroup–specific incidence for LSBE, ensures that the framework is not restricted to Japan and can be directly applied to other regions, including Western countries, by substituting local cost parameters. This generalizability enhances the relevance of the findings beyond Japan and provides a foundation for international comparisons of surveillance strategies.
Fifth, structured risk–stratified surveillance will become increasingly important for maintaining early–stage EAC detection while avoiding unnecessary procedures in low–risk groups as opportunities for incidental detection decline in Japan [3] due to changes in gastric cancer screening practices and health checkup patterns. Policymakers may consider incorporating these findings into future national guidelines to optimize resource allocation and improve patient outcomes.
Finally, the findings suggest that Japan’s upper gastrointestinal screening system and post–Helicobacter pylori eradication surveillance could be restructured to incorporate EAC risk, integrating BE/EAC surveillance into existing endoscopic pathways [3]. Such integration may improve efficiency in lower–risk populations with rising EAC incidence while maintaining appropriate monitoring for higher–risk patients.
5.Conclusions
This incidence–based evaluation demonstrates that the value of surveillance for Barrett’s esophagus varies substantially across subgroups defined by segment length and dysplasia grade. Annual endoscopy is cost–effective for HGD, whereas annual breath test is cost–effective for LGD. In contrast, surveillance offers limited value for NDBE. These findings support prioritizing surveillance resources for higher–risk groups while avoiding unnecessary procedures in low–risk populations.
By grounding the model in empirically observed cancer incidence and avoiding assumptions about unobservable progression states, this study provides a transparent and reproducible framework for evaluating surveillance strategies. Its incidence–based hybrid structure makes the framework broadly generalizable and applicable beyond Japan, including to Western healthcare systems.
Emerging technologies, including AI–assisted endoscopy, sponge test, breath test, and miRNA test, may further enhance the cost–effectiveness of surveillance, particularly in intermediate– and high–risk groups. The development of non–endoscopic procedures is also expected to reduce patient burden and expand access to surveillance.
As EAC incidence is projected to rise in Japan, these findings offer timely, population–specific evidence to guide future surveillance recommendations. Adoption of dysplasia–stratified, incidence–aligned strategies may help optimize resource allocation while maximizing clinical benefit in Japan and could serve as a model for other regions with similarly low but increasing EAC incidence.
Supporting information
Data Availability
All data used in this study were obtained from publicly available sources and are fully described within the manuscript.
Acknowledgments
None.
Funding
This research received no specific grant from any funding agency in the public, commercial, or not for profit sectors.
Conflict of Interest
The author declares no conflicts of interest.
Ethics Statement
This study used publicly available, aggregated national statistics and did not involve human subjects; therefore, ethical approval was not required.
List of Abbreviations
- AI
- Artificial intelligence
- BE
- Barrett’s esophagus
- EAC
- Esophageal adenocarcinoma
- HGD
- High-grade dysplasia
- ICER
- Incremental cost-effectiveness ratio
- LGD
- Low-grade dysplasia
- LSBE
- Long-segment Barrett’s esophagus
- miRNA
- MicroRNA
- NDBE
- Nondysplastic Barrett’s esophagus
- NMB
- Net monetary benefit
- QALY
- Quality-adjusted life-year
- SSBE
- Short-segment Barrett’s esophagus
- USSBE
- Ultrashort-segment Barrett’s esophagus