A comparison of three methods for integrating historical information for Bayesian model averaged benchmark dose estimation

▸ Incompatible historical data have different impact on current BMD estimates depending on integration methods. ▸ The pooled data analysis has evident impact but limited applicability. ▸ The Bayesian hierarchical model is statistically and biologically valid and can have mild impact. ▸ The power prior method can hardly have impact if the data are highly incompatible.