This review provides a vista of the current opportunities and remaining challenges in the area of in vitro-in vivo extrapolation, with particular emphasis on drug binding terms in predictive models, which has been the source of much controversy. Although the importance of fu(inc) (fraction unbound in in vitro incubations) has been acknowledged for decades, it is not always applied in practice. This is somewhat disappointing, since although it may be onerous to measure this term for large numbers of compounds, algorithms to estimate the term from logD(7.4) or logP have been detailed in the literature. These are sufficiently robust to negate routine measurement in early drug discovery. Several groups have recently established convincing relationships between unbound in vivo and in vitro metabolic intrinsic clearance (CL(int)). In the authors' laboratory, correlations of this type have been constructed for rat, dog and Man. The use and interpretation of these models within a drug discovery setting is discussed. The quantitative prediction of drug-drug interactions from in vitro cytochrome P450 (CYP) inhibition data remains a challenge. Although extensive literature databases are at last emerging, apparent ad hoc use of terms for in vivo inhibitor concentrations and only occasional consideration of fu(inc) may only have confused matters. The effect of accounting for drug binding on the accuracy of predictions is reviewed. Other themes including the impact of fu(inc) on relative activity factors (RAFs) and how in vitro data quality and inter-laboratory differences can confound quantitative human pharmacokinetic predictions are also developed.