We derive a new density-biased Monte Carlo technique which preserves detailed balance and improves the convergence of grand-canonical simulations of a species with a strong preference for an interfacial region as compared to the bulk. This density-biasing technique is applied to the solubility of "alcohol" molecules in a mesoscopic model of the lipid bilayer, a system which has anesthetic implications but is poorly understood. © 2010 American Institute of Physics.