A model for predicting individuals' absolute risk of esophageal adenocarcinoma: Moving toward tailored screening and prevention.

Abstract

Esophageal adenocarcinoma (EAC) is characterized by rapidly increasing incidence and poor prognosis, stressing the need for preventive and early detection strategies. We used data from a nationwide population-based case-control study, which included 189 incident cases of EAC and 820 age- and sex-matched control participants, from 1995 through 1997 in Sweden. We developed risk prediction models based on unconditional logistic regression. Candidate predictors included established and readily identifiable risk factors for EAC. The performance of model was assessed by the area under receiver operating characteristic curve (AUC) with cross-validation. The final model could explain 94% of all case patients with EAC (94% population attributable risk) and included terms for gastro-esophageal reflux symptoms or use of antireflux medication, body mass index (BMI), tobacco smoking, duration of living with a partner, previous diagnoses of esophagitis and diaphragmatic hernia and previous surgery for esophagitis, diaphragmatic hernia or severe reflux or gastric or duodenal ulcer. The AUC was 0.84 (95% confidence interval [CI] 0.81-0.87) and slightly lower after cross-validation. A simpler model, based only on reflux symptoms or use of antireflux medication, BMI and tobacco smoking could explain 91% of the case patients with EAC and had an AUC of 0.82 (95% CI 0.78-0.85). These EAC prediction models showed good discriminative accuracy, but need to be validated in other populations. These models have the potential for future use in identifying individuals with high absolute risk of EAC in the population, who may be considered for endoscopic screening and targeted prevention.