Background: The incidence of esophageal adenocarcinoma has recently increased in Asia, including Japan. A system to identify individuals at high risk for Barrett's esophagus (BE), a pre-cancerous condition of esophageal adenocarcinoma, among the general population is needed to perform endoscopic surveillance appropriately. We therefore developed risk prediction scores for BE at health checkups in Japan.
Methods: 4128 consecutive health checkup examinees were retrospectively enrolled from October 2021 to March 2022. A prediction score for BE was developed based on the linear transformation of β-regression coefficients in a multivariable regression model incorporating BE predictors. Internal validation was performed by evaluating discrimination and calibration of the prediction model.
Results: Three prediction scores corresponding to BE based on its length were developed: all lengths, ≥ 1 cm, ≥ 2 cm. All scores were internally validated, and the model calibration was excellent. The performance of the prediction models was better for longer BE, with a c-statistic of 0.70 for BE ≥ 2 cm, than for shorter values. The prediction score for BE ≥ 2 cm yielded sensitivity and specificity of 52.9% and 78.6% in high-risk subjects and 91.2% and 29.3% in intermediate- or high-risk subjects, respectively.
Conclusions: This prediction score can potentially increase the endoscopic detection of BE by identifying potentially high-risk individuals from the general population. This is the first report on developing a prediction score for BE that may suit the Japanese population.