Design: nationwide registry-based cohort study with endoscopy history reconstructed from pathology reports
Population: 8,914 Dutch patients treated curatively for Barrett-related dysplasia or oesophageal adenocarcinoma, 2008-2018
Primary outcome: proportion diagnosed with versus without prior endoscopic surveillance
Effect: 90% (95% CI 89-91) presented de novo; stage II or above in 70% de novo vs 11% under surveillance

Every patient in the Netherlands treated curatively for Barrett-related dysplasia or oesophageal adenocarcinoma between 2008 and 2018 was identified from two national registries - 8,914 patients - and their endoscopy history reconstructed from pathology reports.

Ninety per cent had no prior endoscopy documenting non-dysplastic Barrett oesophagus in the one to five years before treatment. Stage was correspondingly worse in that group: 70% had stage II disease or above, against 11% of those who had been under surveillance. Among the 1,989 patients who received organ-preserving endoscopic treatment, 65% had still presented de novo, and among those needing surgery or chemoradiotherapy, 97% had.

The finding is not that surveillance fails. Among the 889 patients who had prior endoscopies, only 11% had advanced neoplasia missed - the programme performed. The problem is upstream: the disease is overwhelmingly diagnosed in people who were never enrolled, because Barrett oesophagus is asymptomatic and most of it is never found. That reframes the question from surveillance interval to case finding, which in Indian practice is a different problem again: oesophageal squamous cell carcinoma predominates over adenocarcinoma, Barrett prevalence is lower, and a surveillance programme built on Western epidemiology would be poorly targeted.

Most Barrett-related cancer presents in people never known to have Barrett oesophagus.
Surveillance performed well for those enrolled - only 11% had missed advanced neoplasia.
Stage at treatment was far worse outside surveillance: 70% stage II or above against 11%.
Two-thirds of patients still eligible for endoscopic therapy had never been under surveillance.
Indian oesophageal cancer is predominantly squamous, so this case-finding argument transfers only partly.

Why it matters

It moves the argument about Barrett surveillance from how often to scope those enrolled to how anyone gets enrolled at all.

The statistics, in plain English

The 90% figure has a tight confidence interval because the cohort is a national census rather than a sample. What it cannot tell you is how many people in the population have undiagnosed Barrett oesophagus, because the denominator is patients who developed cancer, not people at risk. So this quantifies the failure of case finding without quantifying how much screening would be needed to fix it.