Regional deposition of inhaled medicines is a valuable metric of effectiveness in drug delivery applications to the lung. In silico methods are now emerging as a valuable tool for the detailed description of localized deposition in the respiratory airways. In this context, there is a need to minimize the computational cost of high-fidelity numerical approaches. Motivated by this need, the present study is designed to assess the role of the extrathoracic airways in determining regional deposition in the upper bronchial airways. Three mouth-throat geometries, with significantly different geometric and filtering characteristics, are merged onto the same tracheobronchial tree that extends to generation 8, and Large Eddy Simulations are carried out at steady inhalation flowrates of 30 and 60 L/min. At both flowrates, large flow field differences in the extrathoracic airways across the three geometries largely die out below the main bifurcation. Importantly, localized deposition fractions are found to remain practically identical for particles with aerodynamic diameters of up to d(p) = 4 mu m and d(p) = 2.5 mu m at 30 and 60 L/min, respectively. For larger particles, differences in the localized deposition fractions are shown to be mainly due to variations in the mouth-throat filtering rather than upstream flow effects or differences in the local flow field. Deposition efficiencies in the individual airway segments exhibit strong correlations across the three geometries, for all particle sizes. The results suggest that accurate predictions of regional deposition in the tracheobronchial airways can therefore be obtained if the particle size distribution that escapes filtering in the mouth-throat (ex-cast dose) of a particular patient is known or can be estimated. These findings open the prospect for significant reductions in the computational expense, especially in the context of in silico population studies, where the aerosol size distribution and precomputed flow field from standardized mouth-throat models could be used with large numbers of tracheobronchial trees available in chest-CT databases.