External validation of a model to determine risk of progression of Barrett's oesophagus to neoplasia

BackgroundA risk prediction model containing sex, smoking history, Barrett's oesophagus length and presence of low-grade dysplasia was found to identify individuals at a higher risk of progression to oesophageal adenocarcinoma or high-grade dysplasia. AimTo externally validate the model predicting risk of progression from Barrett's oesophagus to neoplasia and assess the predictive utility of additional factors. MethodsWe conducted a retrospective cohort study among individuals from the population-based Northern Ireland Barrett's register with a histologically confirmed diagnosis of Barrett's oesophagus (with intestinal metaplasia) between 1993 and 2005. The association between a points based model and risk of progression to high-grade dysplasia or oesophageal adenocarcinoma until 2010 was assessed using Cox Proportional Hazards model. Model performance was assessed using area under the receiver operating characteristics curves (AUROC), sensitivity and specificity. ResultsWe identified 1198 individuals with Barrett's oesophagus of whom 54 progressed. The model discriminated reasonably well between progressors and nonprogressors, with an AUROC of 0.70 (95% CI 0.63-0.78). When categorised into low, intermediate and high risk groups, the AUROC was 0.68 (95% CI 0.61-0.74). Compared to using data on dysplasia and segment length for risk stratification, the model resulted in a net reclassification improvement of 20.9%. ConclusionsThis external validation provides further evidence that a model based on sex, smoking, Barrett's segment length and baseline low-grade dysplasia may help to risk stratify patients after an initial diagnosis of Barrett's oesophagus. The model also performed better than the use of low-grade dysplasia status alone for risk-stratification.