Screening for Barrett’s Oesophagus: Are We Ready for it?

Oesophageal adenocarcinoma (OAC) is a poor prognosis cancer [1] that is often preceded by histologically defined premalignant lesions called Barrett’s oesophagus (BO). This provides excellent opportunities for screening (identification of the disease) and surveillance (disease monitoring). Pre-emptive screening and surveillance programmes could have a big impact on outcomes, survival and comorbidity associated with treatment, since outpatient-based endoscopic therapies for premalignant lesions can prevent invasive disease. To date the mainstay for diagnosis and monitoring has been endoscopic-based technology [2]; however, the relative expensive and invasive natures of oesophago-gastro-duodenoscopy (OGD) impede the effective implementation of using this diagnostic modality in a population-wide screening programme. This has become even more apparent with the pressures on diagnostic services during the coronavirus pandemic. These limitations have thus warranted development of other less invasive, more affordable screening tools for mass screening as well as for surveillance and diagnostic triage.

The precursor lesion for OAC is non-dysplastic BO (NDBO) which can subsequently progress to low-grade dysplasia (LGD), high-grade dysplasia (HGD) and eventually to OAC [3]. The annual risk of progression to OAC steadily increases from 0.1 to 0.5% in NDBO to 6% or more in HGD [4‐7]. In some cases, progression can be rapid without opportunity to diagnose intermediate dysplastic steps. However, since population-based studies show that BO diagnosis prior to OAC detection decreases disease mortality [8], and since there has been such an improvement in outpatient, endoscopic therapies for early disease, there is a strong rationale for programmes to improve detection of BO.

A recent systematic review and meta-analysis of papers from the past 10 years reported that 11.8% of OAC cases had a known diagnosis of BO [9••], even taking into account increasing endoscopy volumes [6, 10‐12]. There are several likely explanations behind the substantial numbers of missed screening opportunities for patients with undiagnosed BO. The societal screening guidelines are recommendations, but national public health agencies have not enforced OGD-based mass screening owing to the poor cost benefits and potential morbidity associated with using endoscopy as a screening tool for a relatively uncommon disease. This is especially the case when compared to the higher prevalence of other cancer types like breast and colon. Although there is consensus on the role of BO as a risk factor for OAC and the need for some form of screening, there are different views on “who”, “when” and “how” to screen. Furthermore, these guidelines are not always supported by high-quality evidence and are rather based on expert opinions [13]. Moreover, the cost efficacy and improved mortality rate of this targeted screening approach has never been evaluated.

In terms of who to screen, guidelines from all societies thus restrict screening to more targeted high-risk populations [2, 14‐18]. Both genetic and non-genetic risk factors have been identified for OAC. However, polygenic risk scores are not yet being applied clinically and family history is used as a proxy measure of inherited genetic risk [19]. Regarding non-genetic factors, chronic and frequent gastroesophageal reflux disease (GORD) is the most significant and well-studied. A landmark population-based case-control study demonstrated that patients with GORD had a significantly greater risk of developing OAC as compared with asymptomatic individuals with an odds ratio of 7.7 (95% CI 5.3–11.4) in patients with recurrent reflux symptoms and 43.5 (95% CI 18.3–103.5) amongst patients with chronic and long-standing symptoms (>20years) [20]. Historically, chronic GORD has thus been the sole symptom used for initiating entry into BO/OAC endoscopic screening programmes. However, approximately 40% of patients with OAC do not present with chronic reflux symptoms despite having severe GORD [21]. This so-called “silent reflux” is thought to be explained by a reduction in oesophageal sensitivity to acid refluxate following metaplastic conversion to BO [22]. Nevertheless, using GORD as a screening criterion has its indisputable merits as modelling on the adult US population indicates that screening test would be most impactful in symptomatic GORD patients since they are responsible for 52% of all cancer cases and necessitate screening only 20% of the population [23]. In this scenario, the term “diagnostic test” may be a more fitting word to use than “screening” since only symptomatic persons are being tested unless individuals are invited pro-actively dependent on a search of prescribing databases for example.

The cost benefits of targeted screening have led to the modification of recommendations in current society guidelines to consider additional risk factors like family history, central adiposity, Caucasian ethnicity and male sex [19, 24‐26]. There is also concern about the risks of over-diagnosis and over-treatment since the use of symptoms and risk factors is a blunt tool to predict BO and is difficult to implement, and it remains the case that the majority of patients diagnosed will not progress to OAC. Hence, combinatorial risk models are required taking into account the feasibility of applying an algorithm in clinical practise.

An alternative two-stage screening approach has been sought whereby a subset of patients identified as having an increased probability of having BO are screened in a primary care setting with a less expensive and less invasive test and then referred for endoscopy confirmation. Statistical risk prediction models using widely available and easy to obtain symptoms and risk factors have been developed to estimate the absolute risk that an individual has BO or would develop OAC [27]. For example, a BO prediction model based on electronic health records (EHR) data including gastroesophageal reflux disease, sex, body mass index and ever-smoker status was shown to identify BO patients with a modest accuracy reporting an AUC of 0.71 (95% CI 0.64–0.77) [28]. With risk prediction tools, there is an added potential opportunity for self-assessment whereby a patient can produce a personalised risk profile for BO using a web-based application. Despite the accessibility-related benefits and cost-effectiveness, none of these models is ideal having reported AUCs ranging from 0.61 to 0.75 [27, 29]. Rosenfeld et al. recently developed and validated a machine learning-based risk prediction model for BO (called MARK-BE) using a comprehensive panel of 8 features observed to be significantly associated with increased risk of BO (age, sex, cigarette smoking, waist circumference, frequency of stomach pain, duration of heartburn and acidic taste and taking acid suppressants) [30•]. The MARK-BE model appears to be more robust than other BO risk prediction algorithms published previously due to the large datasets used in the study and the cross validation performed. The model was trained internally with a dataset (collected from BEST 2 case-control screening study (ISRCTN 12730505)—training dataset (n = 776), testing dataset (n = 523)) and subsequently validated externally in an independent dataset (collected from the BOOST case-control study (ISRCTN 58235785) n = 398) reporting an area under the receiver-operator curve (AUC) of 0.81 (95% CI 0.74–0.84, sensitivity set at 90% and specificity of 58%). Although these results look promising, the findings have to be validated in a primary care setting and the optimal thresholds of risk at which screening is warranted using this model have to be determined.

For the current screening paradigm to be successful in OAC prevention, we need to identify the “at-risk” population with greater precision and/or develop other safe, economically viable, minimally invasive screening techniques that can allow effective, systematic screening for BO. For instance, an attractive strategy of “precision cancer prevention” has been proposed whereby a 5-tier system is used to stratify individuals into more precise risk groups with each group being allocated risk-appropriate screening and management options [23]. For example, individuals belonging to the lowest risk strata would require minimal intervention and simple cost-effective screening tools whereas high-risk groups would be eligible for more precise but also more invasive and expensive techniques (Fig. 1). Such a comprehensive and tailored screening protocol would not only incorporate a wider population base, including individuals who are dismissed by current guidelines, but also leverage the use of confirmatory endoscopy only in situations where it is warranted thus saving resources.

As noted, although white light endoscopy remains the cornerstone of the current BO screening practise, it is unlikely to be suitable for mass screening and even routine screening of at-risk individuals because of the direct and indirect cost burden associated with sedation [31] and side effects [32], albeit a highly safe and generally well-tolerated procedure. Trans-nasal endoscopy (TNE) is an alternative endoscopic technique that uses ultra-thin endoscopes (outer diameter <6mm) performed in the sitting position with the instrument inserted through the nose for endoscopic imaging of the distal oesophagus. TNE reduces gagging and vomiting risk thus improving patient tolerability making it more suited for implementation in primary care [33‐35]. It has also been used in mobile research vans for oesophageal assessment and screening for BO in the community [36]. In expert hands, TNE is as sensitive and specific as standard OGD in detecting BO in at-risk population [37]. Result from a meta-analysis of 34 studies with 6659 patients showed a comparable technical success rate (94.0% and 97.8%, respectively) from both TNE and OGD with a significantly higher proportion of patients preferring the former method [38]. TNE is thus recommended in recent guidelines for BO screening due to its comparable diagnostic accuracy, improved patient tolerability and lower direct and indirect cost. This technique has however not been widely embraced, and this may reflect its limited availability, since it requires investment in specialised equipment, decontamination facilities and trained operators, as well as attitudinal barriers amongst clinicians and patients. Further modifications have been carried out on the original TNE to overcome these limitations. The EG Scan TM II (Intromedic Ltd., Seoul, South Korea) is a second-generation TNE device comprising a disposable trans-nasal video esophagoscope with a compact image processing unit. In a tertiary population, with an increased BO prevalence, the EG Scan has been shown to be feasible, safe and capable of detecting BO of any length with a sensitivity value of 0.90 (95% CI, 0.83–0.96) and a specificity value of 0.91 (95% CI, 0.82–0.96), in comparison to conventional esophagogastroduodenoscopy [39]. It also had a high acceptability rate amongst the 89% of participants in the study. Of note, TNE devices have little to no biopsy sampling capabilities and cannot be used for endoscopic therapy. Since biopsy-based diagnosis currently holds the most accurate clinical diagnostic value for BO and OAC, such tools may thus only be useful in two-step screening programme where they could be used for preliminary triage testing in order to enrich the population and then subject individuals with suspected BO to conventional endoscopy for acquisition of multiple, larger, surveillance biopsies to detect dysplasia and enable risk stratification and treatment.

Oesophageal capsule endoscopy, an alternative to the TNE technology, comprises a wireless pill-sized capsule enclosed with a camera, battery and radio transmitter. Images are transmitted by the radio to a digital receiver worn on the waist of the patient and subsequently downloaded into a computer (approximately 8 hrs later) for analysis. Like TNE, OCE does not require sedation and can be carried out in an office setting. The advantage of OCE over conventional TNE is that it is a single-use device obviating the need for equipment reprocessing. The traditional OCE technology has however been reported to have suboptimal diagnostic accuracy with rapid oesophageal transit times which could hinder thorough inspection of the oesophagus [40]. Therefore, newer versions of OCE have been developed so called wireless capsule endoscopy in efforts to overcome problems of rapid oesophageal transit and to allow prolonged imaging and inspection of suspected areas. An example of such innovative modification is the detachable magnetically controlled capsule endoscopy [41]. Tethered capsule endomicroscopy is another variant of OCE that uses optical frequency domain imaging technology [42] to capture 3-dimensional, cross-sectional microscopic images of the entire oesophagus. Subtle tissue microscopic subsurface information that can otherwise be missed by endoscopy is thus provided without requiring microscopic assessment of sampled biopsies. Studies on TCE are promising as the method has been shown to be safe, capable of distinguishing patients with and without BO and feasible in primary care setting for screening BO [43, 44].

Despite great strides made with the OCE technology, larger prospective studies in the relevant primary care population are needed to assess cost-effectiveness and the diagnostic accuracy of these methods for the identification and surveillance of preneoplastic or neoplastic oesophageal lesions.

An alternative strategy to the use of imaging for BO screening would be to sample cells along the oesophagus using minimally invasive non-endoscopic devices and subsequent analysis of molecular aberrations and/or cytopathologic features associated with BO on the cells collected. Non-endoscopic cytological methods have historically been investigated as screening and early detection tools for oesophageal squamous cell cancer in high prevalence countries particularly China. Although these minimally invasive techniques such as brush cytology, mesh and balloon samplers were safe in the majority of study subjects with minimal adverse effects reported, none of these historic studies lead to implementation of the methods established in population-based screening programme owing to poor test sensitivity of traditional cytology. This led the Fitzgerald lab to develop an alternative minimally invasive cell collection device that would retrieve a higher number of cells in an operator independent manner and to augment cytological examination with biomarkers. Although this approach has not yet been recommended in current society guidelines, the recent technical advances and randomised trial data are moving towards clinical implementation.

The main non-endoscopic cell collection devices currently at different stages of development include the Cytosponge (Covidien products, Medtronic, Minneapolis), EsophaCap (Capnostics, Doylestown) and EsoCheck (Lucid Diagnostics, New York) with each technology having its specific strengths and limitations (Table 1). These devices were developed with the intent to be used for large-scale triage testing in the diagnostic or screening setting, and as such, the procedure is performed by a trained nurse or physician, and a positive result test would warrant a follow-up confirmatory diagnosis via the standard OGD method.

The Cytosponge comprises a 30-mm polyurethane sponge, compressed within a gelatine capsule about the size of a vitamin pill attached to a string. Once the patient swallows this capsule and it reaches the gastric cardia, the tightly packed capsule opens up to form a sponge. Cells lining the entire length of the oesophagus and oropharynx are collected when the Cytosponge is retrieved by pulling on the attached string. Similar to the Cytosponge, the EsophaCap is also a sponge on string device although slightly smaller and softer than the Cytosponge but likewise comprises a swallowable encapsulated sponge attached to a tether. The EsoCheck on the other hand is a swallowable balloon-based oesophageal sampling device comprising of a collapsible encapsulated balloon attached to a thin silicone catheter which is connected to a syringe. When the operator identifies the pressure change of the lower oesophageal sphincter the balloon is inflated by injecting air via the catheter, the balloon is gently withdrawn into the distal oesophagus sampling cells from 5–6cm above the GOJ. The device is designed to selectively target the distal oesophagus for sampling as subsequent deflation of the balloon through the catheter results in its inversion back into the capsule thus protecting the acquired bio-sample from further dilution or potential contamination by proximal oesophagus during the capsule retrieval process through the mouth. This more targeted sampling with EsoCheck increases the signal-to-background-noise ratio. However, this feature makes the sample acquisition process more operator dependent. The Cytosponge mesh has a more abrasive surface than the balloon thus allowing deeper samples to be collected. Oesophageal cytology cells captured by these devices are analysed using biomarkers ascertained for their sensitivity and specificity for BO and with the potential to add biomarkers for dysplasia detection.

Although samples collected via the Cytosponge can be assayed for several biomarkers including methylation [53], multigene next-generation sequencing panels [54], and microRNAs [55] useful for screening and surveillance, the current modus operandi for detecting BO from specimens obtained using this device is immunohistochemical staining of Trefoil Factor Protein 3 (TFF3) which is an indicator of intestinal metaplasia (the histopathological feature of BO). The slide-based technique permits the identification of gastric columnar cells as a quality control metric to ensure that the sampling device reached the stomach. In addition, other morphological features can be assessed including atypia indicative of dysplasia and inflammation, as well as application of additional biomarkers for p53 aberration and an immune-histochemical surrogate for copy number [52, 56]. The other two cell sampling technologies use BO-related methylated DNA biomarkers (MDMs), which unlike immunohistochemical staining, do not require subjective interpretation by a pathologist since MDMs can be scalable for automated high throughput testing. Moreover, a methylation-based circulating biomarker, (Epi proColon, Epigenomics) was approved by the U.S. Food and Drug Administration for colorectal cancer screening in 2016 [57].

Alternative non-invasive approaches for BO detection include the use of liquid biopsy whereby BO or early cancer-related biomarkers are assessed via an individual’s peripheral blood or breath samples. Blood or breath sampling is highly feasible and minimally invasive and is therefore more likely to improve acceptability and tolerability of BO screening. However, the most pressing challenge with these approaches is the ability to achieve accurate sensitivity and specificity whilst minimising false positives as well as the need to depend on further tests to confirm BO/early cancer diagnosis.

Current screening guidelines for BO recommend against mass screening since the standard method for BO diagnosis and monitoring, OGD is invasive and expensive thus making it an inappropriate screening tool for a relatively uncommon cancer like OAC. This is even a more pressing issue considering that the ongoing pandemic has resulted in limited access to diagnostic services.

Most guidelines advocate for targeted screening approach whereby individuals who are most at risk of developing BO are subjected to endoscopic screening and then surveillance if warranted. However, this approach is not impacting on the high mortality from OAC. In order to improve this situation, a new clinical pathway designed to systematically identify the at-risk population for BO has to be set in place. Furthermore, cost-effective, less invasive screening tools are needed to improve accessibility to screening opportunities. Indeed, great strides are being made in these two areas. Cost-effective, easily accessible risk prediction algorithms have for instance been tested as a modality for identifying and enriching the population for BO screening. Although the majority of algorithms developed thus far have reported fair AUC’s, they may still be capable of serving enrichment purposes when used in adjunct with other more accurate screening tools. More prospective evaluation of such algorithms is required.

As regards endoscopy alternatives, newer generation of imaging techniques like TCE and OCE have undergone a series of modifications to improve portability and reduce operator dependence increasing their practicality for mass screening. Considering that biopsy sampling is limited with these technologies, in-depth assessment of the diagnostic accuracy of these tests in primary care setting is warranted. Non-endoscopic cell collection device coupled with biomarker analysis is another promising area with randomised controlled trial level evidence to support the utility of Cytoposnge-TFF3 to identify BO.

Combining all the different areas of screening discussed thus far, a three-tier precision cancer prevention-based screening programme whereby (i) easily accessible BO risk assessment algorithms are used for population-wide first-stage screening, (ii) more accurate, cost-effective and minimally invasive tools like TNE/OCE, non-endoscopic cell collection devices or liquid biopsies are used to further triage test the at-risk population in primary care setting, and (iii) suspected cases of BO or early cancer are subsequently followed up using standard endoscopy, could improve the effectiveness of the screening process. By following such a regimen, screening would be made accessible to a larger pool of individuals in an affordable manner thus reducing the number of BO cases that are missed by current practises.

With the ongoing debate on the benefit of BO surveillance, there needs to be studies evaluating the impact of BO screening using these new technologies on overall disease mortality. Furthermore, the low progression rate of BO as well as associated psychological stress amongst many other negative consequences of BO diagnosis reinforce the need for efforts to continuously be made on developing robust biomarker panel that can risk stratify BO patients. Only when this is reached would we be able to unlock the full potential of BO screening for improved clinical outcomes of OAC.