Introduction
Screening for Barrett’s esophagus (BE) could lead to earlier detection and improved survival of esophageal adenocarcinoma (EAC). In the Barrett’s Esophagus Screening Trial 3 (BEST3), an offer of the Cytosponge-Trefoil Factor 3 (TFF3) test led to 10x more BE diagnoses among a reflux population taking acid-suppressant medications. Utilising BEST3 data, we 1) estimate the proportion of missed BE or stage-1 EAC (BE/EAC) under the current primary care reflux referral pathway, 2) define a target screening population, and 3) calculate the trade-off between the enrichment criteria and the proportion of BE/EAC diagnosed.

Methods
The BEST3 study was a multicentre, pragmatic, randomised controlled trial across 109 general practices in England.¹ In this post-hoc analysis, we used the intervention (n=6834) and control arm (n=6388) to estimate the number of missed BE/EAC under the current reflux referral pathway (aim 1), determined which clinical risk factors predict a diagnosis of BE/EAC among the 1654 individuals who swallowed the Cytosponge, and modeled the probability of a diagnosis of BE/EAC (aim 2). We then used different probabilities cut-off (≥3%, ≥5%, and ≥10%) of a BE/EAC diagnosis to estimate the starting age to screen and the trade-off between the population to screen and the proportion of BE/EAC that would be diagnosed using population data from the Office of National Statistics (ONS) of England as a reference (aim 3). We set an upper age limit of 74 years for practical considerations.

Results
In the control arm receiving usual care, the prevalence of BE/EAC was 0.2% (13/6388), compared to 10.6% (680/6388) in the Cytosponge arm, suggesting approximately 98% of BE/EAC cases may be undetected under current care. Risk factors significantly associated with BE/EAC diagnosis were male sex and age. Using male sex and age, we plotted the probability of a diagnosis of BE/EAC by age and by selecting a ≥3%, ≥5%, and ≥10% probability of BE/EAC diagnosis, we showed that the starting age to screen for BE/EAC at these cut-offs were 50, 55 and 65 for males, and 60, 65 and no screening for females (Table 1). Estimating the change in population-to-screen when going from a ≥3% to ≥5%, and to ≥10% cut-off showed that a 5% probability threshold is likely the optimal strategy as it lowers the population to screen by 45.2% (population of males aged 55-74 and females 65-74) while still diagnosing 71.0% of BE/EAC cases compared to the reference population of screening all males and females from age 50-74. The number needed-to-screen to detect one BE/EAC was 10.7.

Conclusions
A targeted screening strategy has the potential to substantially improve detection rates for BE and stage-1 EAC compared with current care. These findings could inform the optimal age for screening males versus females to maximise the yield of screening according to the resources available.

Peter Sasieni

Appearances