Variations in the exposure to risk factors may be used to explain past cancer trends and to predict its future burden. This study aimed to develop a model to describe and predict the variation of esophageal cancer incidence in 1995-2005, taking into account changes in exposures to risk factors. We adapted an existing model to calculate the expected variation in the number of esophageal cancer cases, between 1995 and 2005, in Australia, Japan, Italy, Portugal, the UK, and the USA, because of changes in exposures to risk factors, taking into account the corresponding lag times. Analyses were based on country-specific data of cancer incidence and exposures to risk factors. We computed 95% credibility intervals through Monte Carlo simulation methods. Absolute deviations between the number of cases predicted and those observed in 2005 ranged between 1.8% in Japan and 23.6% in the UK among men and 0.0% in Japan and 18.0% in Australia among women. In Italy and Japan, deviations did not exceed 3%. The UK registered the worst model performance. The variation in esophageal cancer incidence was mainly influenced by changes in fruit and red meat intake, and BMI. For nearly half of the sex-specific and histological type-specific predictions performed, credibility intervals included the observed number of cases. This study proposes a framework for the analysis of the contribution of changes in exposure to different factors to esophageal cancer incidence trends and for long-term predictions at a population level.