A scoring model encompassing just four traits accurately predicted which patients with Barrett’s esophagus were most likely to develop high-grade dysplasia or esophageal adenocarcinoma, researchers reported in the April issue of Gastroenterology (2017 Dec 19.).
Those risk factors included sex, smoking, length of Barrett’s esophagus, and the presence of baseline low-grade dysplasia, said, of Swedish Medical Center, Seattle, and her associates. For example, a male with a history of smoking found to have a 5-cm, nondysplastic Barrett’s esophagus on histology during his index endoscopy would fall into the model’s intermediate risk category, with a 0.7% annual risk of progression to high-grade dysplasia or esophageal adenocarcinoma, they explained. “This model has the potential to complement molecular biomarker panels currently in development,” they wrote.
Barrett’s esophagus increases the risk of esophageal adenocarcinoma by anywhere from 30 to 125 times, a range that reflects the multifactorial nature of progression and the hypothesis that not all patients with Barrett’s esophagus should undergo the same frequency of endoscopic surveillance, said the researchers. To incorporate predictors of progression into a single model, they analyzed prospective data from nearly 3,000 patients with Barrett’s esophagus who were followed for a median of 6 years at five centers in the United States and one center in the Netherlands. At baseline, patients were an average of 55 years old (standard deviation, 20 years), 84% were men, 88% were white, and the average Barrett’s esophagus length was 3.7 cm (SD, 3.2 cm).
The researchers created the model by starting with many demographic and clinical candidate variables and then using backward selection to eliminate those that did not predict progression with a P value of .05 or less. This is the same method used in the Framingham Heart Study, they noted. In all, 154 (6%) patients with Barrett’s esophagus developed high-grade dysplasia or esophageal adenocarcinoma, with an annual progression rate of about 1%. The significant predictors of progression included male sex, smoking, length of Barrett’s esophagus, and low-grade dysplasia at baseline. A model that included only these four variables distinguished progressors from nonprogressors with a c statistic of 0.76 (95% confidence interval, 0.72 to 0.80; P less than .001). Using 30% of patients as an internal validation cohort, the model’s calibration slope was 0.99 and its calibration intercept was -0.09 cohort (perfectly calibrated models have a slope of 1.0 and an intercept of 0.0).