Dutch, UK and US professionals’ perceptions of screening for Barrett’s esophagus and esophageal adenocarcinoma: a concept mapping study

Health care professionals participated in digital concept mapping. This is a standardized mixed-methods approach that integrates qualitative procedures to generate a wide array of ideas and quantitative methods to organize these ideas and expose interconnections and core areas [17]. In contrast to focus groups or surveys, the concept mapping process is participant-driven and involves stakeholders directly in the interpretation of the results. Ethics approval was acquired from the regional ethics committee CMO Arnhem-Nijmegen in the Netherlands (Ref no. 2021–7354). All participants provided informed consent before the start of the study.

We purposefully selected professionals that were involved in esophageal cancer care or specialized in cancer screening, including clinicians, researchers, and policy makers/advisors. We aimed for diversity in practice setting (e.g., public health, primary care, or hospital), and geographical region. We invited professionals who were part of the professional network of members of the research team (RF, SW, JK, MB, PS). We sent personal invitations to Dutch and American professionals and used pre-existing infrastructures (i.e., the BEST3 and Oesophageal Cancer Clinical and Molecular Stratification consortia) to send a general invitation to relevant British professionals. For representation of policy makers’ perspectives, we invited professionals working on BE or EAC screening guidelines (US/UK) and professionals involved in established cancer screening programs (NL). Invitations were sent by e-mail until at least 15 professionals per country agreed to participate, since a sample size of 10 to 40 participants is recommended for this method [17].

Data were collected digitally from March 2021 to October 2021 in three phases: brainstorming, sorting, and rating (Fig. 1). For brainstorming, participants were presented with the following ‘focus prompt’ by e-mail: ‘Generate statements on your perceptions (e.g., perceived barriers, facilitators, advantages, disadvantages, implications or worries) of Barrett’s esophagus and esophageal adenocarcinoma screening’. Participants were instructed to individually brainstorm for ten minutes and to write down concise statements that contained a single idea or thought and submit them to the researcher by e-mail. Spelling or grammatical errors were corrected, and duplicate statements were removed (by JS). A second researcher (YP, MB, or PS) was consulted in case of uncertainty. The unique verbatim statements from each country were subsequently uploaded onto three separate web-based platforms for the Netherlands, UK and US using Concept Systems software version 4.0175, Concept Systems, Inc. (Ithaca, NY). All respondents were provided with a URL that directed them to the platform of their country, e.g. each British participant was presented with the statements generated by all other British participants. Participants were first instructed to sort the statements into groups based on perceived thematic similarity, and to provide a label for each group (see Additional file 1 for the specific wording of the instructions). If a participant sorted statements based on value (e.g., ‘important’ vs ‘not important’) or failed to sort at least 95% of the statements, their data were removed from the analysis (to facilitate the identification of meaningful relationships among statements and ensure comprehensive task engagement) [17]. In the second task of the same survey, participants rated all statements on an eleven-point Likert-type scale using the following question: ‘When evaluating the acceptability of Barrett's esophagus and esophageal adenocarcinoma screening, how relevant is this statement or question?’ (ranging from 0 = not at all relevant to 10 = very relevant). Data from participants that rated at least five statements were included, based on the requirement settings of the concept mapping software.

We applied multidimensional scaling and hierarchical cluster analysis to create cluster-rating maps using the Concept Systems software (see Additional file 1 for a detailed description). The cluster-rating maps provide a visual representation of the participants' sorting data with each point representing a statement and the distance between each point reflecting how frequently the statements were sorted together by participants. The map also depicts mean rating values, relative to the other clusters, as stacked layers [17]. We applied pattern match analysis to compare the mean relevance ratings of each cluster between two subgroups (i.e., ‘clinicians’ vs ‘researchers and policy makers’). Pearson r coefficient (ranging from 0 to 1) was used to describe correlations between the mean ratings. A low correlation indicates differences in priority-setting between subgroups.

Interpretation of the final cluster-rating maps was conducted in digital meetings with the representative team members from each country (by JS, LR, YP, RF, SW, JK, MB, PS). To provide an integrated overview of Dutch, British, and American participants’ perspectives and to explore similarities and differences, one researcher (JS) colour coded similar clusters across the three countries. The maps were subsequently discussed (by JS, LR, YP, MB, PS) to come to a final categorization of each cluster in overarching themes which are outlined in Table 1.

A total of 67 health care professionals participated in the brainstorm activity. Figure 1 shows participant retention rates across the different concept mapping steps (see Additional file 2 for information on non-responders). Sorting data from seven participants were excluded for analysis due to failure to sort at least 95% of the statements (n = 3), or not sorting based on thematic similarity (n = 4). Table 2 shows participants’ characteristics across the three countries, with the majority being male (64.0%) working in various health care settings with over 10 years of experience in their field.

In response to the focus prompt, Dutch professionals generated a total of 123 unique statements, British professionals 106, and American professionals 101. The final list of statements and clusters is available in Additional file 3, stratified by country and sorted by mean cluster/statement relevance rating.

Eight core themes were identified after comparing and integrating the findings on the Dutch, British and American cluster rating maps: (1) Benefits, (2) Harms, (3) Clinical effectiveness concerns, (4) Screening population, (5) Screening modality, (6) Resources, (7) Ownership, and (8) Public communication (Table 1).

Figures 5, 6 and 7 shows a comparison of the rating patterns of clinicians versus other professionals. Dutch clinicians, in evaluating the relevance of statements for the acceptability of BE and EAC screening, assigned relatively higher scores to statements within the categories ‘Potential health benefits’, ‘Capacity’ and ‘Target population identification’. In contrast, Dutch researchers and policy makers assigned higher relevance scores to statements within the themes ‘Legitimacy’ and ‘Harms of screening’ (with mean cluster ratings ranging from 5.46 to 6.51). Clinicians and other professionals had similar rating patterns in the UK and US data.

Professionals emphasized in their statements within the categories ‘Benefits’ and ‘Clinical effectiveness’ that the potential of screening to prevent EAC and reduce EAC-related mortality is currently theoretical and lacks empirical evidence. This finding is consistent with a recent American survey involving 315 gastroenterologists and primary care providers, where the majority of participants expressed the need for more data evaluating both the benefits (69% concurred) and potential harms (62% concurred) of BE screening, including insights from RCTs (84% concurred) [12]. The deficiency in supporting evidence likely contributes to professionals’ low belief in the effectiveness of BE and EAC screening, as reported in earlier surveys [12‐14]. These perspectives are understandable, given that the impact of screening on EAC-related mortality remains unstudied [6]. Robust evidence will be required in order for healthcare professionals to accept BE or EAC screening. To the best of our knowledge, there is one ongoing RCT, BEST4, aiming to assess whether offering the Cytosponge-TFF3 to patients on medication for heartburn symptoms could lead to an improvement in EAC-associated mortality [18]. However, results are not expected until 2035.

Participants’ statements regarding the prerequisites for an acceptable screening modality were largely congruent with our prior analysis of expert opinion articles, namely: minimally-invasive and acceptable to the public, high sensitivity and specificity, easily administered, and low costs [19]. Evaluation of the statements on an item-level reveals that most professionals perceived the Cytosponge-TFF3 and breath test as more acceptable modalities for large-scale screening compared to conventional endoscopic screening. This sentiment broadly aligns with the general public’s perspective, though some individuals express reservations about the anticipated discomfort associated with the Cytosponge-TFF3 [20, 21]. Notably, none of the professionals suggested the use of transnasal endoscopy as a screening modality, reflecting its limited clinical adoption since being endorsed as an alternative for conventional endoscopy by the American College of Gastroenterology in 2016 [6, 22]. The public also appears to perceive transnasal endoscopy as the least preferable test modality [23].

Our study findings resonate with previous surveys showing that GPs perceive time constraints and difficulty identifying suitable candidates for BE and EAC screening as barriers [12, 13, 16]. A prior American survey reported that 58% of GPs encountered challenges in determining who should undergo upper endoscopy for BE screening, compared to only 16% of gastroenterologists [12]. One potential explanation is that the published literature on risk factors and screening for EAC has predominantly been confined to gastroenterology journals. Furthermore, the underrepresentation of GPs in our study (see Additional file 2) might suggest a lower inclination towards conducting BE or EAC screening, or perhaps a perception that it falls outside their designated scope of practice. In this light, gaining a deeper understanding of GPs’ perspectives on their role in shared decision-making regarding BE and EAC screening, as well as their involvement in administrating screening tests, is crucial. This is especially true in the context of implementing a primary care-based tool like the Cytosponge-TFF3. We therefore recommend conducting an additional qualitative study focused on GPs’ role preferences, information needs and facility requirements.

The divergence in terminology between American participants, who referred to the public communication category as ‘public awareness,’ and British participants, who labeled it as ‘patient education,’ likely reflects country-specific approaches to initiating cancer screening. In Europe, screening policies typically adhere to a publicly coordinated and systematic approach, incorporating proactive screening invitations and reminders that necessitate balanced educational materials [24]. In the US, screening practices adopt a more decentralized model relying on individuals taking the initiative. Individual healthcare providers play a pivotal role in determining the most appropriate screening tests and engaging in shared decision-making. This distinction may explain why American professionals suggested strategies such as using social media platforms and incorporating labels on over-the-counter acid-suppressants to enhance public awareness of BE and EAC.

Part of the British participants were connected to the previously described BEST3 trial [9]. This real-world experience provided them with valuable insights into practical barriers and facilitators of BE screening. These encompassed securing funding, ensuring easy and equitable community access, developing high-quality public communication materials, training personnel in the Cytosponge-TFF3 procedure, and ensuring adequate pathology and endoscopy services for the follow-up of screening participants. Consequently, fostering collaboration with the UK to draw from their experiences and potentially launching a multinational screening trial could prove valuable for the Netherlands. This is because the Netherlands has a similar approach to cancer screening but limited experience with large-scale BE or EAC screening trials in primary care settings [10, 11].

A major strength of our study using a systematic combination of qualitative and quantitative methods is that we engaged participants in the analytic stages of the study. This minimized individual interpretation bias. Additionally, the purposeful selection of professionals allowed us to collect input across different organizational levels and profession types in a wide range of geographic areas. Nonetheless, some limitations should also be considered. First, there is a risk of participation bias, i.e., those who did not take part may have different views which are not represented. British responders may have had a relatively high interest in the topic since they were either BEST3 collaborators [9] or connected to esophageal cancer consortia. Second, the general public’s perspective was not within the scope of this study and has been covered elsewhere [21]. Third, although we removed sorting data that was not categorized based on thematic similarity, some clusters still contained statements that did not optimally match the assigned label (e.g., the statement ‘There is no clear evidence on surgical prevention of reflux’ was grouped in the cluster ‘Patient education’). It is conceivable that participants who generated relatively few categories negatively impacted the quality of the data in terms of thematic similarity, by grouping statements together which are reasonably dissimilar in content [17]. Fourth, concept mapping is a mixed-methods approach primarily used to generate insights, it is not intended to test null-hypotheses regarding the relevance of themes. Average cluster ratings should therefore be interpreted in this light.