Human HIV integrase inhibitors are a novel class of antiretroviral drugs that act by blocking incorporation of the proviral DNA into the host cell genome, a crucial step in the life cycle of HIV. In the present work, quantitative methods for prediction of human pharmacokinetics were used to guide the selection of development candidates from a series of dihydroxypyrimidine and N-methylpyrimidinone carboxamide inhibitors of HIV integrase, which are cleared mainly by O-glucuronidation. The pharmacokinetics of 10 drugs from this series was determined in several preclinical species, including rats, dogs, rhesus monkeys, and rabbits, and the in vitro turnover, plasma protein binding, and blood/plasma partition ratio were studied using preparations from both preclinical species and humans. Two clearance prediction methods, based on physiologically based scaling or allometric scaling normalized for differences in microsomal turnover, were used to extrapolate human clearance. For three clinical candidates, including the novel AIDS drug raltegravir (MK-0518, Isentress), oral drug exposure was predicted and compared with that observed in healthy human volunteers. Both scaling methods gave a reasonable correspondence between predicted and observed oral exposure. Prediction errors for the physiologically based method were less than 1.7-fold for two drugs, including raltegravir, and less than 3.5-fold for one drug. The exposures predicted using normalized allometric scaling were within 1.1- to 1.5-fold of observed values for all three compounds. The accuracy of prediction by normalized allometric scaling was similar when using data from either four preclinical species or from rats and dogs only. The prediction methods used may be applicable to other drugs cleared predominantly by glucuronidation.