Predicting absolute risks for oesophageal adenocarcinoma (#161)
Background & Aims: Adenocarcinoma of the oesophagus (OAC) is a rapidly fatal disease. We aimed to develop a prediction model to estimate the absolute 5-year risks of developing OAC for people with different profiles of risk factors.
Methods: We derived the risk model using epidemiologic data from 364 patients with incident OAC and 1580 population controls. Significant risk factors were fitted into a multivariate unconditional logistic regression model. The final multivariate model was combined with age- and sex-specific OAC incidence data to estimate absolute 5-year risks for OAC. We performed a 10-fold cross validation of the data to assess the relative performance of the model.
Results: The final risk model included terms for highest level of education, body mass index, smoking status, frequency of gastroesophageal reflux symptoms and/or use of acid suppressant medications, and frequency of use of non-steroidal anti-inflammatory drugs. The population attributable risk for the model was 0.92. A 10-fold cross validation produced an area under the receiver operator characteristic curve (AUC) statistics of 0.75 (95% confidence interval, 0.66-0.84), indicating good discrimination. Adding alarm symptoms, frequency of symptoms of dysphagia and unexplained weight loss to the model significantly improved discrimination (AUC=0.85, 95%CI 0.78-0.91).
Conclusion: Risk models offer the potential to identify people at higher than average risks of OAC, who may benefit from targeted cancer prevention strategies aimed to reduce premature mortality from this disease.