WATS^3D INTEROBSERVER STUDY SHOWS EXCELLENT REPRODUCIBILITY AMONG ACADEMIC GI PATHOLOGISTS

Introduction:
Barrett’s esophagus (BE), a pre-malignant precursor of esophageal adenocarcinoma (EAC), when recognized early through screening, could potentially allow for endoscopic surveillance or endoscopic eradication therapy. As such, major GI societies recommend targeted screening for BE based on age, race, obesity, smoking, chronic gastroesophageal reflux disease (GERD) and family history of BE/EAC. However, at-risk population is defined variably across societal recommendations, which has led to ambiguity in defining the potential population eligible for BE screening.

Aim:
To estimate the national burden of BE screen-eligible population in accordance with GI societal recommendations.

Methods:
We used the nationally representative 2012 National Health Interview Survey (NHIS) to estimate the U.S. national burden of BE screening based on guidance documents from four major societies (ACG Guidelines 2022, AGA Clinical Practice Update 2022, ASGE Guidelines 2019, BSG Guidelines 2014). We also estimated the number of screen-eligible patients based on different combinations of risk factors in the entire U.S. adult population. Analyses accounted for complex sampling of NHIS.

Results:
The non-institutionalized population in the U.S. in 2012 was 234,920,670. Of these, 43.6% were age ≥50y, 48.1% were males, 76.1% were non-Hispanic white, 39.9% were ever-smokers (17.9% were current smokers), and 27.3% were obese. Overall, 22.7% reported GERD symptoms in the prior year. Based on GI societies’ BE screening recommendations, the screen-eligible population ranged from 19.7 million (8.4%) based on BSG 2014 recommendations to 120.1 million (51.1%) based on AGA Clinical Practice Update 2022 (Table 1). There was wide variation in the number of risk factors in the screen-eligible population, ranging from 78.1% with ≥2 risk factors to 24.7% with ≥4 risk factors. GERD patients had fewer other risk factors: 19.6% with ≥2 risk factors, and 5.4% with ≥4 risk factors (Figure 1).

Conclusions:
The estimated burden of BE screen-eligible population based on current societal recommendations is high and variable, potentially imposing a considerable resource burden, especially if endoscopic screening is pursued. Better understanding of risks and benefits of screening, precise risk stratification and high-performing, scalable, non-invasive screening modalities are needed to improve appropriate targeting and reduce the burden on patients and resources associated with population-based BE screening.

Background-Patients with short segment Barrett’s oesophagus (BO) and no intestinal metaplasia (IM) may inadvertently end up on surveillance programmes and contribute to an unnecessary burden on healthcare resources, interventional risk, patient anxiety and environmental footprint. In the UK BO is defined by the presence of IM on histology and the presence of visible columnar epithelium above the gastroesophageal junction (GOJ) at gastroscopy. National guidelines recommend surveillance is offered to those with maximal BO length (M)>= 3cm. A Cytosponge service to triage patients on Barrett’s surveillance to gastroscopy and avoid diagnostic delay was established in 2020 to address COVID19 delays. The surrogate biomarker for IM on Cytosponge cytology is trefoil factor (TFF3).
Methods-All patients on BO surveillance programme due between 30/11/2020 and 30/11/2022 were offered Cytosponge. Exclusions were previous dysplasia of any grade, previous fundoplication, strictures, pregnancy and patient preference. 2 research nurses and 3 clinical nurse specialists were trained in delivering Cytosponge. Results were relayed to patients within 4 weeks of the procedure. Patients with dysplasia/atypia/p53 or TFF3+ve are described elsewhere. Those with negative or equivocal TFF3 were identified and their previous gastroscopy and histology reports were reviewed for the presence of visible BO (Prague criteria) and IM respectively. Patients with M <3cm and no previous IM were discharged from surveillance following consultation with the specialist nurse, advice and safety netting. Figure 1 shows the decision tree used for managing patients with a negative TFF3.
Results- Over the 2y period 313 patients had a successful Cytosponge with adequate cytology sampling. 232 (74.2%) patients had a positive TFF3. 81 (25.8%) had negative or equivocal TTF3. Of these, at last gastroscopy 24 (7.7%) had Barrett’s length reported to be 3cm or greater and continue on surveillance programme with next gastroscopy in 12-18 months. 57 had Barrett’s length of <3cm or unspecified length reported at last gastroscopy. 22 of these had M<3cm without IM at previous gastroscopy and were discharged without further investigation. The remaining 35 had or await repeat gastroscopy and to date 36/313 (11.5% of the programme) were discharged.
Conclusion- An additional unexpected benefit from the Barrett’s surveillance Cytosponge service was to identify at least 11.5% patients not requiring ongoing surveillance. This has a positive impact to patients, healthcare systems and helps reduce the healthcare carbon footprint.

Background: WATS^3D (CDx Diagnostics, Suffern NY) is a diagnostic platform that utilizes mucosal brush sampling combined with neural network technology to enhance detection of dysplasia in Barrett’s Esophagus (BE) when used as an adjunct to forceps biopsies (FB). Diagnostic reproducibility among pathologists evaluating FB has been suboptimal. The purpose of this study was to evaluate interobserver variability in diagnosing BE-associated dysplasia in WATS^3D brush specimens among academic GI pathologists previously inexperienced with this technology.

Methods: After a 4 hour combined in person/virtual training session of WATS^3D diagnostic criteria, 5 academic GI pathologists with no prior experience with this technology assessed WATS^3D brush samples from 60 BE patients. Each of the 60 brush samples was processed into one formalin-fixed paraffin-embedded (FFPE) H and E-stained slide and one PAP-stained smear slide (120 slides in total), all of which were scanned digitally for pathology evaluation. Study pathologists were asked to assess the images for either non-dysplastic BE (NDBE), Indefinite for dysplasia (Indef), low (LGD) or high grade dysplasia/carcinoma (HGD/CA) from each of the two digitalized slides separately, and also when combined into a final worst diagnosis per patient. All WATS^3D cases selected for this study had their diagnosis (20 with NDBE, 20 with LGD, 20 with HGD/CA) confirmed in the FB that was obtained at the same endoscopic procedure by 1 pathologist who did not participate in the interobserver analysis. Kappa statistics were performed for diagnostic variability.

Results: Overall, all 5 pathologists agreed on the final patient diagnosis in 88.3% of cases (≥4/5 agreed in 98.3%). The Kappa value for overall worst diagnosis was 0.93 (CI=0.87-0.98), which is considered excellent (Kappa=0.93 for FFPE slides, 0.97 for smear slides). The Kappa values based on specific diagnoses were as follows: NDBE=0.94, LGD=0.90, HGD/CA=0.95. Only one pathologist diagnosed Indef, and this was applied in only two cases. When the analysis was based only on the presence or absence of dysplasia (either LGD or HGD/CA), all 5 pathologists agreed on the patient’s final diagnosis in 95.0% of cases (Kappa=0.95), and this value reached 100% in the smear slide specifically. Finally, there was 100% agreement between the WATS^3D dysplasia diagnoses rendered by the majority of study pathologists and the FB diagnoses.

Conclusions: Interobserver agreement of WATS^3D specimens for BE-associated dysplasia is extremely high among non-industry GI pathologists without any prior cytology or WATS^3D training. The results are considerably higher than prior studies based on FB alone. Most notably, determination of dysplasia vs no dysplasia was nearly perfect. After focused training, WATS^3D can be utilized in non-industry settings with excellent diagnostic accuracy.