An extended set of multidrug-resistance modulators of the propafenone type were investigated using CoMFA and CoMSIA. A number of 3D-QSAR models were derived from steric, electrostatic, and hydrophobic fields and their combinations. The hydrophobic fields alone and in combination with the steric and both (steric and electrostatic) fields yielded the models with the highest cross-validated predictivity, in agreement with a previous analysis of a smaller data set of propafenone-type multidrug-resistance (MDR) modulators. Inclusion of lipophilicity did not lead to an improvement of the models. The results point to the importance of hydrophobicity as a space-directed molecular property for MDR-modulating activity. The influence of variable selection applying the GOLPE procedure was investigated with an external test set. Variable-selection procedure was repetitively applied, keeping at each stage variables with uncertain contribution to the models. For the CoMFA-based 3D-QSAR models, an increase in external prediction quality was found. In contrast, the CoMSIA-based 3D-QSAR models were not improved by the GOLPE variable-selection procedure.