Simulation within the grand canonical ensemble is the method of choice foraccurate studies of first order vapour-liquid phase transitions in modelfluids. Such simulations typically employ sampling that is biased with respectto the overall number density in order to overcome the free energy barrierassociated with mixed phase states. However, at low temperature and for largesystem size, this approach suffers a drastic slowing down in samplingefficiency. The culprits are geometrically induced transitions (stemming fromthe periodic boundary conditions) which involve changes in droplet shape fromsphere to cylinder and cylinder to slab. Since the overall number densitydoesn't discriminate sufficiently between these shapes, it fails as an orderparameter for biasing through the transitions. Here we report two approaches toameliorating these difficulties. The first introduces a droplet shape basedorder parameter that generates a transition path from vapour to slab states forwhich spherical and cylindrical droplet are suppressed. The second simplybiases with respect to the number density in a tetragonal subvolume of thesystem. Compared to the standard approach, both methods offer improvedsampling, allowing estimates of coexistence parameters and vapor-liquid surfacetension for larger system sizes and lower temperatures.