Cost‐effective identification of Barrett's esophagus in the community: A first step towards screening

Abstract

Background and Aim

The first step towards developing a screening strategy for Barrett's esophagus (BE) is the identification of individuals in the community. Currently available tools include endoscopy, less‐invasive non‐endoscopic devices, and non‐invasive risk stratification models. We evaluated the cost of potential strategies for identification of BE as a first step towards screening.

Methods

Two hypothetical cohorts of the general population aged ≥ 50 years with BE prevalence rates of 1.9% and 6.8% were modeled. Four potential screening tools were evaluated: (i) risk stratification based on non‐weighted clinical factors according to US/European guidelines, (ii) weighted risk stratification using algorithmic models, (iii) less‐invasive devices such as Cytosponge + trefoil factor 3 (TFF3), and (iv) endoscopy. Using a decision‐analytic model, the cost per BE case identified and the cost‐effectiveness were compared for six potential BE screening strategies based on combinations of the four screening tools; (i) + (iv), (ii) + (iv), (iii) + (iv), (i) + (iii) + (iv), (ii) + (iii) + (iv), and only (iv).

Results

The cost per BE case identified was lowest for the weighted risk stratification followed by Cytosponge‐TFF3 then endoscopy strategy at both 1.9% and 6.8% BE prevalences (US$9282 and US$3406, respectively) although it was sensitive to the cost of less‐invasive devices. This strategy was also most cost‐effective for a BE prevalence of 1.9%. At BE prevalence of 6.8%, the Cytosponge‐TFF3 followed by endoscopy strategy was most cost‐effective.

Conclusions

Incorporating weighted risk stratification and less‐invasive devices such as Cytosponge‐TFF3 into BE screening strategies has a potential to cost‐effectively identify BE in the community although device cost and the community prevalence of BE will impact the optimal strategy.

Introduction

Esophageal adenocarcinoma (EAC) has rapidly increased in incidence in Western countries. ¹ As many patients are diagnosed with late stage cancer, potentially curative treatment is possible in only half of cases, and overall, 5‐year survival is < 20%. ¹ , ² Conversely, the 5‐year survival rate of early asymptomatic cancers is > 80%. ² Shifting the stage at presentation from late to early offers a significant opportunity to improve outcomes at the population level through early endoscopic treatment pathways. Barrett's esophagus (BE) is the known precursor for EAC, and screening for BE followed by endoscopy surveillance offers an opportunity for early stage EAC detection. ³

For BE identification, conventional sedated endoscopy remains the gold standard. However, endoscopy is invasive, costly, ⁴ and is not ideal for wide‐scale screening of the general population, and for this reason, guidelines considering likely cost‐effectiveness in a screening scenario only recommend its use for high‐risk patients. ⁵ , ⁶ , ⁷ , ⁸ , ⁹ , ¹⁰ The US and European guidelines propose clinical risk factor‐based criteria to stratify and focus resources on patients at high risk of BE. ⁵ , ⁶ , ⁷ , ⁸ , ⁹ , ¹⁰ However, these criteria have reported low sensitivities (35%–40%) for identifying BE. ¹¹

For that reason, other tools for screening have been explored, and less‐invasive potential devices have been described, including non‐endoscopic swallowable cell collection devices, blood tests, and breath tests. ¹² , ¹³ , ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ Among these, the Cytosponge + trefoil factor 3 (Cytosponge‐TFF3) test has been most widely evaluated for cost‐effectiveness. ¹⁸ , ¹⁹ , ²⁰ In addition, cheaper non‐invasive screening tools have also been considered, including risk stratification based on clinical factors. Algorithmic models that weigh risk factors have achieved higher performance than guideline criteria in stratifying individuals according to the probability of having BE. ¹¹ , ²¹ Most recently, electronic health record‐based machine learning models have also been reported to offer valid predictions of BE cases. ²² , ²³ , ²⁴ Given their non‐invasive nature and low cost, algorithmic models and machine learning models have potential as initial tools in a screening pathway. Overall, a variety of screening strategies with diverse options are now available, but the cost‐effectiveness of these strategies is poorly understood.

For screening to be cost‐effective, an appropriate strategy for the identification of BE in the community needs to be applied, followed by ongoing cost‐effective surveillance of the identified BE cases. The aim of this study was to identify a cost‐effective strategy for the identification of BE as a first step towards a screening pathway. We hypothesized that the sequential use of different strategies, starting with risk stratification models followed by less‐invasive devices and then confirmation of BE using endoscopy might be the most cost‐effective approach to identify BE in the community. To test this, we evaluated the cost‐effectiveness of multiple strategies for BE identification, which combined risk stratification, less‐invasive screening tools, and endoscopy.

Methods

This study followed the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) reporting guidelines. ²⁵ The analysis used the Australian health care system perspective. All costs are presented in 2024 US dollars (AUS$1 = US$0.66, https://www.google.com/finance/quote/AUD‐USD, accessed on June 17, 2024). As the aim of this study was to determine the cost of identifying a case of BE in the community and not the cost of subsequent surveillance, the time horizon was restricted to the BE diagnostic pathway, starting with the initial screening modality and ending with the diagnosis of BE confirmed by endoscopy and biopsy. As a result, costs and effectiveness were not discounted. Modeling and analysis were conducted using TreeAge Pro Healthcare, version 2023 R2.0 (TreeAge LLC).

Study population

The analysis was conducted using two hypothetical cohorts of the general population aged 50 years and over: Australian and USA. The BE prevalence considered two scenarios based on published studies; (i) 1.9% based on Australian and Swedish population studies ²⁶ , ²⁷ or (ii) 6.8% based on US population studies. ²⁸ , ²⁹ Separate analyses that considered the different prevalences of BE in male and female cohorts in both populations were also evaluated.

Screening tools for BE

A systematic literature review was conducted to identify BE screening tools and their sensitivity and specificity in the different populations. Detailed results from this review are the subject of a separate publication (under preparation). Multiple BE screening tools were identified in the review and classified into four broad categories: (i) risk stratification based on non‐weighted clinical factors, (ii) risk stratification based on weighted clinical factors, (iii) less‐invasive devices, and (iv) sedated endoscopy (Table S1).

Risk stratification based on non‐weighted clinical factors

Guideline‐based prediction models for BE were categorized as risk stratification based on non‐weighted clinical factors when the stratification in the guidelines was based simply on summing the number of clinical risk factors. ¹¹ As the cost of these models is generally very low and the accuracy is moderate, they were considered for use as an initial first screening tool, before using less‐invasive tools and/or endoscopy for confirmation of BE. The American College of Gastroenterology (ACG) guideline‐based stratification was applied in the base‐case analysis. ¹⁰ This model had the highest area under the receiver operating characteristic curve (ROC‐AUC) for this strategy, although the diagnostic accuracy was broadly similar for all four guideline‐based models (Table S1). As no published validation of the updated 2022 ACG guideline criteria for the general population was available, the data from the 2016 ACG Guideline validation were used. ¹¹

Risk stratification based on weighted clinical factors

Several stratification models have considered the influence of each clinical factor and weighted them to determine risk. ¹¹ , ²¹ , ²² , ²³ , ²⁴ , ³⁰ , ³¹ , ³² , ³³ Some of these are based on machine learning calculations. ²² , ²³ , ²⁴ These models are generally low cost but show a higher predictive performance than non‐weighted stratification models (Table S1). The machine learning‐based stratification model reported by Iyer et al. was applied in the base‐case analysis ²² as this model had the highest ROC‐AUC for this strategy (Table S1).

Less‐invasive devices

Less‐invasive devices include three types of swallowed devices, blood‐based circulatory microRNAs test, breath tests, and capsule endoscopy. ¹² , ¹³ , ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ The cost of these tools is generally intermediate to high, and they have moderate to high accuracies and might be considered for use as a screening tool before confirmatory endoscopy. Cytosponge‐TFF3, which is a swallowed device, was applied in the base‐case analysis as it is commercially available and has been widely used in Europe. ¹² , ¹⁸ , ¹⁹ , ²⁰

Sedated endoscopy

We modeled the use of sedated endoscopy as a screening strategy and to confirm BE visually and by biopsy following the above three screening tools. The diagnostic accuracy of biopsies was based on an international prospective study. ³⁴ We excluded un‐sedated trans‐nasal endoscopy as this is not a widely used screening tool in Australia or elsewhere.

Model structure

A decision‐analytic model was constructed to compare the cost‐effectiveness of six potential pathways for BE screening using TreeAge Pro 2023 (Fig. 1). The six strategies compared were as follows:

1. endoscopy‐alone; where the whole population underwent endoscopy;

2. non‐weighted factors‐based risk stratification followed by endoscopy, where endoscopy was performed only in high‐risk populations identified by the non‐weighted stratification. This strategy is recommended in the current US and European guidelines. ⁵ , ⁶ , ⁷ , ⁸ , ⁹

3. weighted factors‐based risk stratification followed by endoscopy, where endoscopy was performed only in high‐risk populations identified by the weighted stratification;

4. less‐invasive devices followed by endoscopy, where endoscopy was performed only in the populations with positive results from less‐invasive devices;

5. non‐weighted factors‐based risk stratification followed by less‐invasive devices and then endoscopy, where less‐invasive devices were applied only in high‐risk populations identified by the non‐weighted stratification, and endoscopy was then performed only in the populations with positive results from less‐invasive devices;

6. weighted factors‐based risk stratification followed by less‐invasive devices and then endoscopy, where less‐invasive devices were applied only in high‐risk populations identified by weighted stratification, and endoscopy was then performed only in the populations with positive results from less‐invasive devices.

A simplified model constructed using TreeAge Pro software is shown in Figure S1.

Figure 1.

Six potential screening strategies modeled.

Input parameters

Test performance

The sensitivity and specificity of the screening tools that were inputted into the model were obtained from the published literature. Systematic reviews were preferred if available, and single studies were used where no recent systematic reviews were available. Where multiple studies were found, the most recent large study was used.

Cost

Cost data were primarily obtained from the Australian Medicare Benefits Schedule (MBS), ³⁵ which is the Australian government health care reimbursement scheme. Cost data for screening tools that were not available in Australia were obtained from the US government reimbursement scheme (Centers for Medicare and Medicaid Services), published literature, or sources referenced in previous cost‐effectiveness studies. To obtain the cost of risk stratification based on clinical risk factors, the fee for general practitioner attendance was applied from the MBS. Input parameters and sources are provided in Table S1.

Study outcomes

The measure of effectiveness was the number of BE cases identified. Identification of BE was defined as the presence of visible columnar mucosa within the esophagus and histologically confirmed intestinal metaplasia from endoscopic biopsy, according to the Australian and US clinical guidelines. ⁵ , ³⁶ The decision‐analytic model computed the number of BE cases identified and the cost per BE case identified for each screening strategy. Because there is no established cost threshold per BE case identified within a screening scenario, the cost threshold target for this analysis was arbitrarily set at US$10 000.

To evaluate the cost‐effectiveness of the six potential screening strategies, the incremental cost‐effectiveness ratio (ICER) or cost needed to find an additional BE case were analyzed. The ICER threshold was set at US$18 500 as suggested by Edney et al., based on the Australian health system. ³⁷ Estimated values from the decision‐analytic model were used to calculate the cost per BE case identified by dividing costs by effects for each strategy and to calculate the ICER by dividing incremental costs by incremental effects for all combinations of six strategies.

Base‐case analysis

In the base‐case analysis, the cost per BE case identified and the ICER were assessed in the two cohorts with the BE prevalence set at 1.9% (Australia/Europe) and then 6.8% (USA). Estimated male and female cohorts for both populations were also evaluated in the sub‐group analyses. The base‐case analysis evaluated all six proposed screening strategies.

Sub‐analysis

A sub‐analysis was performed to explore whether a serum/blood‐based screening tool would also have potential to be cost‐effective. For this, the circulatory microRNAs reported by Bus et al. ¹⁵ were modeled. An alternative serum/blood‐based tool reported by Rubenstein et al. ²¹ was not considered as the addition of these serum biomarkers to the clinical factor‐based risk stratification did not substantially alter the ROC‐AUC compared with the clinical factor‐based stratification alone (0.704 vs. 0.700, P = 0.45).

Sensitivity analyses

First, one‐way sensitivity analyses varying the BE prevalence (1.0%–8.0%) and the cost of less‐invasive devices (US$30–$2000) were performed to examine how each of these variables affected the cost of identifying one case of BE. Second, a tornado analysis was performed to identify the variables that had the greatest influence on the ICER results. Input parameters were varied within plausible ranges (Table 1), reflecting published data available for the various screening tools, including 4 non‐weighted risk stratification tools, 13 weighted risk stratification tools, and 6 less‐invasive devices (Table S1).

Table 1.

Base‐case input parameters, one‐way sensitivity analysis (SA) ranges, and assigned probabilistic sensitivity analysis (PSA) distributions

Parameter	Base case value	One‐way SA range	PSA distribution	SD†	Source
Prevalence of Barrett's esophagus	1.9% (AUS/Europe) or 6.8% (USA)	1.0–8.0%	N/A	N/A	26 , 27 , 28 , 29 High estimate reflecting an article from USA 20
Test sensitivities and specificities
Risk stratification based on non‐weighted factors
Sensitivity	0.39	0.39–0.43	Beta	0.15	11
Specificity	0.77	0.72–0.77	Beta	0.04	11
Risk stratification based on weighted factors
Sensitivity	0.76	0.57–0.90	Beta	0.03	11 , 21 , 22 , 23 , 24 , 30 , 31 , 32 , 33
Specificity	0.76	0.21‐0.78	Beta	0.01	11 , 21 , 22 , 23 , 24 , 30 , 31 , 32 , 33
Less‐invasive devices
Sensitivity	0.8	0.78–0.93	Beta	0.03	12 , 13 , 14 , 15 , 16 , 17
Specificity	0.92	0.73‐0.93	Beta	0.03	12 , 13 , 14 , 15 , 16 , 17
Sedated confirmatory endoscopy
Sensitivity	0.92	0.87–0.97	Beta	0.05	34
Specificity	1	1	N/A	N/A	Barrett's esophagus is confirmed by the histological assessment on endoscopy
Costs (US$)
Risk stratification based on non‐weighted factors	$27.3	± 20%	Gamma	2.7	MBS, Estimates reflecting potential across country variations
Risk stratification based on weighted factors	$27.3	± 20%	Gamma	2.7	MBS, Estimates reflecting potential across country variations
Less‐invasive devices	$174	$30–$2000	Gamma	34.8	MBS, USA Centers for Medicare & Medicaid Services,, 38 Estimates reflecting potential across country variations and devices
Sedated endoscopy	$917	$300–$2000	Gamma	91.7	MBS, Estimates reflecting potential across country variations

Note: AUS$1 = US$0.66 (June 17, 2024).

Abbreviations: AUS, Australia; CI, confidence interval; MBS, Medicare Benefits Schedule; PSA, probabilistic sensitivity analysis; SA, sensitivity analysis; SD, standard deviation; USA, United States of America.

^†SD = based on 95% CI of the base case.

Finally, probabilistic sensitivity analyses were performed by including distributions of plausible ranges for test sensitivities, test specificities, and costs using point estimates and standard deviations applied in the base‐case analysis (Table 1). Beta distributions were applied to test specificities and sensitivities while gamma distributions were applied to the costs. ³⁹ The variables were randomly sampled to calculate ICERs for each input simultaneously within its distribution. A Monte Carlo simulation of 10 000 iterations was performed, allowing for an estimation of the overall uncertainty. The cost‐effectiveness acceptability curve was generated to show the uncertainty across a range of willingness‐to‐pay thresholds.

Results

Base‐case analyses

The mean cost for each case of BE identified was highest in the endoscopy‐alone strategy at US$52 460 and US$14 658, and lowest in the weighted stratification followed by less‐invasive devices and then endoscopy strategy at US$9282 and US$3406, at 1.9% and 6.8% BE prevalences, respectively (Table 2). The endoscopy‐alone strategy was the most effective at identifying BE cases, with 92% sensitivity. The least effective strategy was the non‐weighted stratification followed by less‐invasive devices and endoscopy, identifying 29% of cases at both 1.9% and 6.8% BE prevalences.

Table 2.

Results of the base‐case analysis: cost per case identified, effectiveness, and incremental cost‐effectiveness ratio

Strategy	Cost per one BE case identified (US$)	The number of BE cases identified per 10 000 population (95% CI)	The number of BE cases missed per 10 000 population (95% CI)	Incremental cost‐effectiveness ratio † (US$; i.e., cost needed to find an additional case)
Prevalence of Barrett's esophagus: 1.9% (AUS/Europe)
Risk stratification model with non‐weighted factors followed by less‐invasive devices and then endoscopy	$16 473	55 (54–55)	135 (135–136)	Ref
Risk stratification model with weighted factors followed by less‐invasive devices and then endoscopy	$9282	106 (106–106)	84 (84–84)	$1702	Ref
Risk stratification model with non‐weighted factors followed by endoscopy	$35 351	68 (68–69)	122 (121–122)	$110 867	$‐37 357	Ref
Risk stratification model with weighted factors followed by endoscopy	$19 303	133 (133–133)	57 (57–57)	$21 275	$59 390	$2388	Ref
Less‐invasive devices followed by endoscopy	$18 586	140 (140–140)	50 (50–50)	$19 937	$48 049	$2638	$4955	Ref
Endoscopy alone	$52 460	175 (175–175)	15 (15–15)	$68 780	$119 430	$63 398	$157 456	$187 956
Prevalence of Barrett's esophagus: 6.8% (USA)
Risk stratification model with non‐weighted factors followed by less‐invasive devices and then endoscopy	$5348	195 (193–196)	485 (484–487)	Ref
Risk stratification model with weighted factors followed by less‐invasive devices and then endoscopy	$3406	380 (380–381)	300 (299–300)	$1358	Ref
Risk stratification model with non‐weighted factors followed by endoscopy	$10 172	244 (241–245)	436 (435–439)	$29 468	$‐8700	Ref
Risk stratification model with weighted factors followed by endoscopy	$5885	475 (475–476)	205 (204–205)	$6259	$15 803	$1366	Ref
Less‐invasive devices followed by endoscopy	$5840	500 (499–501)	180 (179–181)	$6154	$13 547	$1718	$4976	Ref
Endoscopy alone	$14 658	626 (625–626)	54 (54–55)	$18 880	$32 111	$17 526	$42 439	$49 932

Abbreviations: AUS, Australia; USA, United States of America; CI, confidence interval; BE, Barrett's esophagus.

^†Incremental cost‐effectiveness ratios are calculated in all combinations between six strategies.

ICERs for the six strategies are reported in Tables 2 and S2. At the ICER threshold of US$18 500, the most cost‐effective strategy at 1.9% BE prevalence was the weighted risk stratification followed by less‐invasive devices and then endoscopy at ICER of US$1702, and at BE prevalence of 6.8%, it was less‐invasive devices followed by endoscopy, with an ICER of US$13 547. In male‐only and female‐only cohorts, the most cost‐effective strategy was the weighted stratification followed by less‐invasive devices and then endoscopy, unless the BE prevalence exceeded 8.2% (male prevalence in the USA ²⁸ ) where the less‐invasive devices followed by endoscopy strategy was the most cost‐effective.

Sub‐analysis

In a sub‐analysis to evaluate the potential for serum biomarker to be cost‐effective, the circulatory microRNAs ¹⁵ followed by endoscopy strategy was absolutely dominated at both BE prevalences of 1.9% and 6.8%, meaning that the strategy cannot be the most cost‐effective. When a combination of circulatory microRNAs and Cytosponge‐TFF3 was tested, a three‐step strategy with the sequential use of these two tools and then endoscopy was also absolutely dominated and not cost‐effective.

One‐way sensitivity analysis for the costs per case of BE identified

One‐way sensitivity analysis to examine the change in cost of identifying one BE case with varying the BE prevalence (1.0%–8.0%) is shown in Figure 2. The weighted risk stratification followed by less‐invasive devices and then endoscopy strategy was the cheapest for identifying BE cases at any BE prevalence. With the reasonable cost threshold per BE case identified set at US$10 000, no strategy was acceptable for a population with a BE prevalence of 1.0%–1.7%, as all strategies cost over US$10 000. With the same threshold, the weighted risk stratification followed by less‐invasive devices and then endoscopy strategy was the most acceptable at a BE prevalence of 1.8%–3.6%, whereas less‐invasive devices followed by endoscopy was the most acceptable at a BE prevalence of 3.7%–8.0%.

Figure 2.

One‐way sensitivity analysis: Cost per one Barrett's esophagus identified with prevalence of Barrett's esophagus. , Endoscopy alone. , Less‐invasive devices followed by endoscopy. , Risk stratification model with weighted factors followed by less‐invasive devices and then endoscopy. , Risk stratification model with weighted factors followed by endoscopy. , Risk stratification model with non‐weighted factors followed by less‐invasive devices and then endoscopy. , Risk stratification model with non‐weighted factors followed by endoscopy.

One‐way sensitivity analysis was also performed to examine the change in cost of identifying one BE case when varying the cost of less‐invasive devices (US$30–$2000) (Fig. S2). At a BE prevalence of 1.9%, the costs per BE case identified were lowest using weighted stratification followed by less‐invasive devices and then endoscopy at a device cost of ≤ US$600 and lowest using weighted stratification followed by endoscopy at a device cost of > US$600. At a BE prevalence of 6.8%, the threshold for device cost was US$517.

Threshold analyses for ICER

Threshold analyses for ICER were performed as tornado analyses (Fig. S3). Based on the above one‐way sensitivity analyses varying the BE prevalence, the strategy of less‐invasive devices followed by endoscopy was most effective with a BE prevalence of 6.8%, and its performance was compared with two strategies: (i) weighted stratification followed by less‐invasive devices and then endoscopy and (ii) weighted stratification followed by endoscopy. These analyses showed that variations in the cost of less‐invasive devices (US$30–$2000) had the greatest influence on the ICER results (US$52 569–$734 675), demonstrating that a decrease in the costs of the less‐invasive devices will substantially lower the ICER. The diagnostic accuracy of weighted stratification was also shown to have a large influence on the ICER.

Probabilistic sensitivity analyses for ICER

The cost‐effectiveness acceptability curve illustrates the probability of strategies being cost‐effective at different ICER or willingness‐to‐pay thresholds (Fig. 3). For the population with a BE prevalence of 1.9%, weighted stratification followed by less‐invasive devices and endoscopy had more than 50% probability of being cost‐effective for finding an additional case of BE across a willingness‐to‐pay threshold range of US$2500–$45 000 (Fig. 3a). For the population with a BE prevalence of 6.8%, the same strategy had more than 85% probability of being cost‐effective across a range of US$2500–$10 000, less‐invasive devices followed by endoscopy had more than 50% probability of being cost‐effective across a range of US$15 000–$47 500, while endoscopy alone had more than 50% chance of being cost effective above US$50 000 (Fig. 3b).

Figure 3.

Cost‐effectiveness acceptability curve based on ICER. (a) At the Barrett's esophagus prevalence of 1.9%. (b) At the Barrett's esophagus prevalence of 6.8%. ICER, incremental cost‐effectiveness ratio; US$, US dollar. , Endoscopy alone. , Less‐invasive devices followed by endoscopy. , Risk stratification model with weighted factors followed by less‐invasive devices and then endoscopy. , Risk stratification model with weighted factors followed by endoscopy. , Risk stratification model with non‐weighted factors followed by less‐invasive devices and then endoscopy. , Risk stratification model with non‐weighted factors followed by endoscopy.

Discussion

This health economic modeling of potential strategies to identify BE in the community revealed that strategies that included less‐invasive devices such as Cytosponge‐TFF3 were generally more cost‐effective than strategies that progressed directly to endoscopy. The cost per case of BE identified was lowest for the strategy that commenced with weighted stratification, then followed by less‐invasive devices and finally endoscopy confirmation. The optimal strategy was influenced by the proposed prevalence of BE, with the most cost‐effective strategy at a BE prevalence of 1.9% being weighted stratification followed by less‐invasive devices and then endoscopy, whereas at a BE prevalence of 6.8%, weighted stratification was not necessary and commencing with less‐invasive devices followed by endoscopy was the more cost‐effective strategy. A caveat to this conclusion was the sensitivity analysis, which suggested that weighted risk stratification followed by endoscopy would also be cost‐effective strategy at higher BE prevalence levels, when the cost of the less‐invasive devices was higher and/or the diagnostic accuracy of weighted stratification was higher.

Current US and European guidelines for BE screening recommend the non‐weighted stratification followed by endoscopy strategy. ⁵ , ⁶ , ⁷ , ⁸ , ⁹ As guidelines are based on published evidence, consensus and expert opinion, they do need to be reconsidered and revised as new evidence emerges. The guideline‐recommended approach was clearly inferior to the proposed weighted stratification followed by less‐invasive devices and then endoscopy strategy from the aspect of both costs and effectiveness in all our simulated populations. This suggests that current screening recommendations should be reconsidered. Consistent with previous studies that reported the cost‐effectiveness of Cytosponge, ¹⁸ , ¹⁹ , ²⁰ the usefulness of less‐invasive devices is further supported by our base‐case analysis which showed that the most cost‐effective strategies included Cytosponge‐TFF3, at all levels of BE prevalence.

However, we should also consider the finding from our sensitivity analyses that the cost of the devices strongly affects cost‐effectiveness and hence the optimal strategy (Fig. S3). In the base‐case analysis, Cytosponge‐TFF3 costs US$174. However, if the cost of less‐invasive devices exceeded US$208 and if the ICER threshold is set at US$18 500, ³⁷ the less‐invasive devices followed by endoscopy approach is less cost‐effective than the weighted risk stratification followed by endoscopy strategy at a BE prevalence of 6.8% (Fig. S3a).

When we evaluated a possible role for serum/blood‐based screening tools using data for the most sensitive and specific tool, circulatory microRNAs reported by Bus et al., ¹⁵ serum/blood‐based screening tool alone or in combination with swallowed devices was not as cost‐effective as the other strategies tested. The previously reported serum biomarkers from Rubenstein et al. also failed to improve cost‐effectiveness, as they added little additional sensitivity or specificity to the clinical factor‐based risk stratification approach, but they did add significant additional cost. ²¹ For serum/blood‐based tools to add value to a screening strategy, they will need to achieve much higher levels of sensitivity and specificity or be able to be delivered at much lower cost.

Our study has identified strategies for the identification of BE in the community. This is the first step for developing a screening pathway. Such a pathway requires the cost‐effective identification of previously unknown BE cases, and then a subsequent strategy for ongoing cost‐effective surveillance. It is the first of these steps that we have addressed in this paper.

Our results suggest that risk stratification based on clinical factors will be an important first step in this pathway, especially when considering a population with lower BE prevalence. For example, at 1.9% BE prevalence, the ICER for the less‐invasive devices followed by endoscopy strategy compared with the weighted stratification followed by less‐invasive devices and then endoscopy strategy was US$48 049, which exceeded the ICER threshold of US$18 500. This confirmed that the three‐step strategy which started with weighted risk stratification is more cost‐effective. Given its non‐invasive nature and lowest costs, the use of weighted risk stratification before less‐invasive devices appears to be a good option.

The cost‐effectiveness of risk stratification has not been considered in current US, European, or Australian guidelines. ⁵ , ⁶ , ⁷ , ⁸ , ⁹ , ³⁶ This probably reflects a lack of evidence when the guidelines were being formulated. It is also important to note that the diagnostic accuracy of risk stratification highly affects cost‐effectiveness. Indeed, this is confirmed by lack of cost‐effectiveness for the strategies which included non‐weighted factors in the scenarios we modeled. Our modeling supports the use of algorithmic models which weight risk factors and not those based simply on summing the number of risk factors per individual. Furthermore, the diagnostic accuracy of algorithmic models was also shown to have a large influence on the ICER (Fig. S3), and this accuracy determines the most cost‐effective strategy (Table S3). Widely validated algorithmic models, however, are scarce, and further investigation and validation is warranted.

It should also be noted that consistent with a previous simulation study by Sami et al., ²⁰ the endoscopy‐alone strategy was not the most cost‐effective, although the strategy was the most effective for identifying BE cases owing to its high sensitivity. If we seek to identify as many BE cases as possible, the endoscopy‐alone strategy is superior. This increases the number of people under future BE surveillance and hence generates further costs for that surveillance. However, screening pathways must take a broader population view and aim for lower costs and most effective use of resources, and to achieve this, the pathway must be tolerant of missing cases.

Our study confirmed that an optimal strategy for identification of BE in the community depends on the prevalence of BE in the population. In populations with a high BE prevalence, universal screening using less‐invasive devices followed by endoscopy might be cost‐effective, whereas universal screening may not be cost‐effective in populations with low BE prevalence. We applied prevalence rates of 1.9% and 6.8% based on published evidence. BE prevalence of 1.9% appears consistent with clinical observation and opinion in most Western countries, whereas prevalence of 6.8% seems high. However, this rate can be supported by previous publications, ²⁸ , ²⁹ therefore it was modeled.

For now, the critical input of the population prevalence of BE is still unclear in most populations. Determining this prevalence is confounded by differences in the definition of BE between different jurisdictions. For example, the presence of intestinal metaplasia is a prerequisite for the BE diagnosis in US and Australian guidelines ⁵ , ⁶ , ⁹ , ³⁶ but not in the British guideline. ⁷ In addition, individuals undergoing endoscopy represent a biased sample and include mainly those with symptoms or a preference for endoscopy as opposed to other diagnostic modalities. ⁴⁰ , ⁴¹ A more accurate BE prevalence would be based on endoscopy with histological diagnosis in a randomly sampled population. However, such studies are rare. For now, more work to determine the prevalence of BE in a population will be needed to determine an optimal screening strategy.

Our study has several limitations. First, the modeling only accounted for the direct medical costs of the BE diagnostic pathway. To fully assess the costs of each pathway, comprehensive costs including out of pocket costs such as over‐the‐counter medication, direct non‐medical costs (e.g., transport or unpaid care), and indirect costs (e.g., absenteeism or presenteeism) should also be applied. However, for the purpose of informing policy decisions on population screening, a health system perspective including only the direct medical costs was considered sufficient as these are what the health care provider pays. Second, the analysis was limited to identifying a case of BE with the number of BE cases identified as the measure of effectiveness. This is not the full screening pathway, and the evaluation of a screening strategy for cost‐effectiveness over a life‐time horizon is beyond the scope of this paper. Our current study did, however, aim to evaluate the cost‐effectiveness of the BE diagnostic process using multiple screening strategies, and therefore this approach should be sufficient to inform the next steps required for the development of cost‐effective screening in the community. The third limitation is that our threshold for cost‐effectiveness was set at US$18 500 based on the Australian health system. ³⁷ This is subject to debate as most thresholds recommended by decision making bodies consider quality adjusted life years (QALY) as a measure of outcome. However, in their paper which defined an Australian threshold, Edney et al confirmed that the base‐case estimate of US$18 500 is sensitive to uncertainty around the estimated morbidity‐related QALY gains, ranging from US$15 058 to $23 990 for the upper and lower bounds of estimated health‐related quality of life improvement. Our probabilistic sensitivity analyses showed the dominance of two strategies including less‐invasive devices across a wide range of willingness‐to‐pay thresholds and confirmed the robustness of our findings. These analyses also demonstrated that the most cost‐effective strategy based on ICER can vary between different countries due to different willingness‐to‐pay thresholds (Fig. 3). For example, if the United States applies its usual threshold of US$50 000, the endoscopy‐alone strategy might be considered cost‐effective at a BE prevalence of 6.8%, but it will not be cost‐effective in countries with lower thresholds such as Australia (threshold of US$18 500) and the United Kingdom (threshold of GBP30 000 = US$39 078 on August 21, 2024). In Australia and the United Kingdom, the strategy of less‐invasive devices followed by endoscopy only becomes cost‐effective at a BE prevalence of 6.8%.

In conclusion, the use of less‐invasive non‐endoscopic devices should be considered for cost‐effective identification of BE in the community as a first step towards developing a screening pathway. The use of weighted risk stratification before less‐invasive devices further improves the cost‐effectiveness of BE identification, especially at a low BE prevalence, whereas non‐weighted risk stratification was not cost‐effective. An optimal screening strategy may not be uniform and should be considered for each population based on the local BE prevalence and the availability and local costs of diagnostic and screening tools.

Supporting information

Figure S1. Decision tree illustrating six potential pathways of Barrett's esophagus screening.

Figure S2. One‐way sensitivity analyses: cost of less‐invasive devices and cost per one Barrett's esophagus identified.

Figure S3. Tornado analyses based on ICER at a Barrett's esophagus prevalence of 6.8%.

Table S1. Screening tools for Barrett's esophagus.

Table S2. Sub‐group analysis in estimated male and female cohorts: cost per case identified, effectiveness, and incremental cost‐effectiveness ratio.

Table S3. Sub‐analysis to evaluate the effect of alternative tools on the cost per case identified and incremental cost‐effectiveness ratio.

Aoki, T. , Watson, D. I. , and Bulamu, N. B. (2024) Cost‐effective identification of Barrett's esophagus in the community: A first step towards screening. Journal of Gastroenterology and Hepatology, 39: 2654–2663. 10.1111/jgh.16762.

Data availability statement

The authors confirm that the data supporting the findings of this study are available within the article and/or its supporting information.