Modelling a population with Barrett's oesophagus from oesophageal adenocarcinoma incidence data

A recent study of adenocarcinoma of the oesophagus (ACO) incidence rates in Denmark showed a steep fall in the over-80 population, interpreted as the result of a decline in the prevalence of Barrett's oesophagus (BO) in this age group, for which three hypotheses were advanced: the specific mortality from ACO and, superimposed, either excess mortality from causes of death unrelated to ACO or a birth cohort effect. The aim of this study was to create models estimating the BO population fitting each of these three hypotheses, in order to select the most plausible hypothesis and to gain insight into the Danish BO population.

Models were designed for these three hypotheses, conforming to the generally accepted 0.4-0.5% annual ACO incidence in BO patients. These models employed expectation-maximization (EM) algorithms, Danish life tables and the observed ACO incidence rates. The models enabled the estimation of a BO population for each hypothesis.

After testing against set criteria, the most plausible model was found to be that describing a birth cohort effect which predicted a +/-5% annual rise in the prevalence of BO and, consequently, in the incidence rate of ACO in Denmark. This prediction was borne out over the subsequent decade.

This rising ACO incidence rate is likely to continue into the foreseeable future. The use of EM algorithms enabled a first estimate of the BO population at risk of ACO, although, owing to the limitations imposed by the models, the age- and gender-specific ACO risk for the entire Danish BO population could not as yet be ascertained.