All half-lives are wrong, but some half-lives are useful.

The half-life of a drug, which expresses a change in concentration in units of time, is perhaps the most easily understood pharmacokinetic parameter and provides a succinct description of many concentration-time profiles. The calculation of a half-life implies a linear, first-order, time-invariant process. No drug perfectly obeys such assumptions, although in practise this is often a valid approximation and provides invaluable quantitative information. Nevertheless, the physiological processes underlying half-life should not be forgotten. The concept of clearance facilitates the interpretation of factors affecting drug elimination, such as enzyme inhibition or renal impairment. Relating clearance to the observed concentration-time profile is not as naturally intuitive as is the case with half-life. As such, these 2 approaches to parameterising a linear pharmacokinetic model should be viewed as complementary rather than alternatives. The interpretation of pharmacokinetic parameters when there are multiple disposition phases is more challenging. Indeed, in any pharmacokinetic model, the half-lives are only one component of the parameters required to specify the concentration-time profile. Furthermore, pharmacokinetic parameters are of little use without a dose history. Other factors influencing the relevance of each disposition phase to clinical end-points must also be considered. In summarising the pharmacokinetics of a drug, statistical aspects of the estimation of a half-life are often overlooked. Half-lives are rarely reported with confidence intervals or measures of variability in the population, and some approaches to this problem are suggested. Half-life is an important summary statistic in pharmacokinetics, but care must be taken to employ it appropriately in the context of dose history and clinically relevant pharmacodynamic end-points.