A series of 28 diarylsulfonylureas with antitumor activity was subjected to a three-dimensional quantitative activity relationship (3D-QSAR) study. Three different QSAR methods, comparative molecular field analysis (CoMFA), hologram QSAR (HQSAR) and comparative molecular similarity indices analysis (CoMSIA), were compared in terms of their potential for predictability. All three QSAR-based models had good predictability and yielded q(2) values 0.74, 0.63 and 0.72, respectively. The CoMFA model provided the highest q(2) and r(2) values, which implied the significance of correlation of steric and electrostatic fields with biological activities. The number of components was 3-4 for all three QSAR methods. The quality of HQSAR or CoMSIA was slightly lower than that of CoNFA in terms of q(2) and r(2) values. HQSAR does not require the generation of a three-dimensional structure of molecules and CoMSIA does not require molecular superposition, therefore they are faster than CoMFA in data processing.