AbstractBackgroundIn a recent coordinated analysis across population‐based cohorts, we have shown that the age‐specific dementia incidence has declined by 13% per decade since 1990. Here, we aim to incorporate these trends into the projection of the future dementia burden, using microsimulation models that can synthesise complex information, take time trends into account, and estimate long‐term and population wide intervention effects.MethodBased on the Microsimulation Screening Analysis (MISCAN) model from cancer research, we developed a dementia microsimulation model. It synthesises data from the population‐based Rotterdam Study with published disease stage durations from several mostly clinical cohorts. We incorporate changes in demographics by defining unique birth and life tables for every 10‐year birth cohort between 1910 and 1980. Three trend scenarios in age‐specific dementia incidence were calibrated to fit the observed incidence data from the Rotterdam Study: 1) no risk differences between birth cohorts; 2) a linear decline of 13% in incidence rate per birth cohort; and 3) two nonlinear declines that result in the same cumulative decline and average incidence rate as the linear trend. Projections of the dementia incidence in the Netherlands until 2050 were calculated for every trend scenario.ResultThe model simulated 10 million individuals, representative of the Dutch population in the 1910 to 1980 birth cohorts. The model was able to fit the observed incidence data well. Figure 1 shows the incidence projections for the Dutch population implementing the aforementioned trend assumptions. Assuming no trend results in the highest incidence projections, the linear trend results in the lowest, and the nonlinear trends lie in between. By 2050, compared to no decline, the linear and nonlinear trends result in 40% and 25‐35% fewer incidence cases.ConclusionAssumptions about the trend in incidence play a crucial role for the projections of the future dementia incidence, and subsequently prevalence and mortality. We developed a model that is well capable of projecting trends on future incidence. An application to the Dutch population projects 25‐40% lower incidence than currently assumed. Modelling the effect of interventions and risk changes on a population level provides essential information for policy makers and researchers.