Loading field map
Pharmacokinetics is the quantitative study of what the body does to a drug. It describes the journey of a drug molecule from the moment it enters the body until it is eliminated, tracing the changes in its concentration over time at various sites. The discipline is built on the premise that the intensity of a drug's therapeutic effect—and its toxicity—is related to its concentration at its site of action. Since that site is usually inaccessible in a living patient, pharmacokinetics uses measurable surrogate concentrations, most often in blood plasma, to model and predict drug behavior. Its clinical counterpart, pharmacodynamics, studies what the drug does to the body; pharmacokinetics provides the exposure side of the exposure–response relationship.
The field addresses four fundamental processes, often summarized by the acronym ADME: absorption (how a drug enters the bloodstream), distribution (how it moves into tissues and organs), metabolism (how it is chemically transformed, primarily in the liver), and excretion (how it and its metabolites leave the body). The central questions are deceptively simple: How much of the drug reaches the systemic circulation? How quickly does it get there? How long does it stay? And how does the answer vary between patients, between doses, and over the course of treatment?
The stakes are high. A drug given at too low a dose may never reach a therapeutic concentration; too high a dose may produce toxicity. Many drugs have a narrow therapeutic window—the range between the minimum effective concentration and the minimum toxic concentration—so small differences in pharmacokinetics can mean the difference between a successful treatment and a serious adverse event. This is why pharmacokinetics underpins dose selection in drug development, the design of dosing regimens in clinical practice, and the regulatory approval of new medicines. It also explains why the same dose of a drug can affect two patients very differently: differences in body size, organ function, genetics, age, and interacting medications all alter the ADME processes.
The roots of pharmacokinetics lie in the early twentieth century, when physiologists and pharmacologists began applying mathematical models to drug disposition. The field's conceptual foundations were laid by researchers who recognized that drug concentrations in the body often decline in a predictable, exponential fashion. The German pharmacologist Hans Horst Meyer and others studied the distribution of drugs between body compartments, but it was not until the 1930s and 1940s that formal compartmental models emerged. The Danish pharmacologist Torsten Teorell published a landmark two-part paper in 1937 that set out a mathematical framework for drug absorption, distribution, and elimination, treating the body as a system of interconnected compartments. His work was largely theoretical and initially had limited practical impact, but it established the core idea that drug disposition could be described by differential equations.
The field matured in the 1950s and 1960s, driven by the development of sensitive analytical techniques—first spectrophotometry, then chromatography—that allowed researchers to measure drug concentrations in biological samples with precision. This period saw the formalization of the two most influential concepts in the field: the compartmental model and the clearance concept. The compartmental approach, associated with researchers such as Sidney Riegelman and John Wagner, treated the body as one or more well-mixed compartments and used exponential equations to describe how drug moves between them. The clearance concept, articulated most influentially by the British pharmacologist Milo Gibaldi and the American pharmacologist Gerhard Levy, reframed drug elimination in terms of the volume of plasma cleared of drug per unit time, a concept borrowed from renal physiology. This shift from rate constants to clearance and volume of distribution proved more physiologically meaningful and more useful clinically.
A second major transformation began in the 1970s and accelerated through the 1980s: the development of population pharmacokinetics. Traditional pharmacokinetic studies involved intensive sampling of a small number of healthy volunteers or patients, yielding precise estimates for each individual. But clinicians needed to know how a drug behaves across a diverse patient population. The American statistician Lewis Sheiner and his colleague Stuart Beal introduced nonlinear mixed-effects modeling, a statistical approach that could analyze sparse, unbalanced data from many patients simultaneously. This allowed researchers to estimate both the typical pharmacokinetic behavior in a population and the magnitude of variability between individuals, and to identify patient characteristics—such as kidney function, body weight, or age—that explain part of that variability.
The field is organized around several distinct but overlapping approaches, each answering a different kind of question.
The classical approach treats the body as a system of compartments. A one-compartment model assumes that the drug distributes instantaneously and uniformly throughout the body, so that the concentration in plasma reflects the concentration everywhere. The decline in concentration after an intravenous dose follows a single exponential decay. A two-compartment model is more realistic: it posits a central compartment (plasma and well-perfused organs) and a peripheral compartment (muscle, fat, and other tissues) into which the drug distributes more slowly. The concentration–time curve then shows a rapid distribution phase followed by a slower elimination phase. Three-compartment models exist for drugs that distribute into deep tissues, but they are less common.
The strength of compartmental models is their mathematical tractability. They yield simple equations that can be fitted to data to derive parameters such as the elimination rate constant, half-life, and volume of distribution. Their weakness is that the compartments are abstract constructs, not anatomical realities. A "peripheral compartment" does not correspond to a specific tissue, and the model cannot predict what happens when physiology changes—for example, in a patient with reduced cardiac output. Compartmental models describe the data well but explain little about the underlying physiology.
A pragmatic alternative, now standard in drug development, is noncompartmental analysis. This approach avoids assuming a specific compartment structure. Instead, it uses model-independent formulas to calculate key parameters directly from the concentration–time data. The area under the concentration–time curve (AUC) is estimated using the trapezoidal rule, and from it one can derive clearance (dose divided by AUC) and volume of distribution at steady state. The terminal half-life is estimated from the slope of the final log-linear portion of the curve.
Noncompartmental analysis is robust because it makes few assumptions, and it is widely used in regulatory submissions for new drugs. However, it is descriptive rather than predictive. It tells you what happened in the studied individuals but cannot simulate what would happen under a different dosing regimen or in a different patient population. It also requires that sampling be continued long enough to capture the terminal phase reliably, which is not always feasible.
A fundamentally different approach, physiologically based pharmacokinetic (PBPK) modeling, builds the model from actual physiology rather than from curve fitting. The body is represented as a network of compartments corresponding to real organs—liver, kidney, lung, muscle, fat—connected by blood flow. Each organ has a real volume and blood flow rate, and drug movement into and out of each organ is governed by the drug's physicochemical properties (such as lipophilicity and protein binding) and by known biochemical processes (such as metabolic enzyme activity). The model is constructed from in vitro data and physiological parameters, then used to predict in vivo behavior.
PBPK modeling has grown dramatically in capability and acceptance since the 2000s, driven by advances in computing power and by regulatory agencies' willingness to consider such models in drug applications. Its great advantage is that it is mechanistic: it can predict how a drug will behave in a patient population that was never studied directly, such as children, pregnant women, or patients with liver disease, by adjusting the physiological parameters. Its limitations are equally real. It requires a large amount of high-quality input data, and its predictions are only as good as the underlying assumptions about organ physiology and drug properties. A PBPK model can be exquisitely detailed and entirely wrong if a key parameter is misestimated.
Population pharmacokinetics is not a separate model of drug disposition but a statistical framework for analyzing data from many individuals. It uses nonlinear mixed-effects models to estimate the typical value of pharmacokinetic parameters in a population, the inter-individual variability around those typical values, and the influence of covariates (patient characteristics) on those parameters. The approach is particularly valuable in clinical settings where only a few blood samples can be taken from each patient, such as in neonates or in critically ill patients.
Population pharmacokinetics has become the standard tool for therapeutic drug monitoring—the practice of measuring drug concentrations in individual patients and adjusting doses accordingly. It also underpins the concept of model-informed precision dosing, in which a patient's individual pharmacokinetic parameters are estimated from their own sparse measurements combined with population priors, allowing the clinician to predict the dose needed to achieve a target concentration. The approach is powerful because it embraces variability rather than ignoring it, but it requires sophisticated statistical expertise and careful model validation.
Contemporary pharmacokinetics is best understood as a layered enterprise in which these approaches coexist and complement one another. In drug development, noncompartmental analysis remains the workhorse for characterizing a new drug's basic pharmacokinetics in early clinical trials. Compartmental models are used when a mechanistic description of the concentration–time profile is needed for simulation. PBPK models are increasingly used to support dose selection, to predict drug–drug interactions, and to extrapolate from healthy adults to special populations. Population pharmacokinetic models are routinely included in regulatory submissions to characterize variability and to support dosing recommendations.
The field has also expanded in scope. Pharmacogenetics and pharmacogenomics have identified genetic variants in drug-metabolizing enzymes (such as the cytochrome P450 family) and transporters that explain a substantial portion of inter-individual variability in drug exposure. This knowledge is now integrated into pharmacokinetic models and, in some cases, into clinical dosing guidelines. Therapeutic drug monitoring has moved from a niche specialty to a standard practice for drugs with narrow therapeutic windows, such as immunosuppressants, certain antibiotics, and anticonvulsants. The rise of biologics—large-molecule drugs such as monoclonal antibodies—has required extensions of classical pharmacokinetic concepts, since these drugs have very long half-lives, are not metabolized by cytochrome P450 enzymes, and may be subject to target-mediated drug disposition, where the drug's binding to its pharmacological target itself influences its clearance.
A notable recent development is the growing use of machine learning in pharmacokinetic modeling. These data-driven approaches can identify patterns in large clinical datasets that traditional models might miss, and they can be used to predict individual drug exposure from routine clinical data. However, they remain complementary to mechanistic modeling rather than a replacement: they require large, high-quality datasets, and their predictions are difficult to interpret mechanistically.
The enduring challenge of pharmacokinetics is the tension between precision and practicality. Mechanistic models offer deep understanding but demand extensive data; empirical models are simple and robust but offer limited insight. The field's trajectory has been toward integrating both—using physiology to inform the structure of models and using data to refine and validate them. The ultimate goal remains unchanged: to predict, with acceptable accuracy, how a particular drug will behave in a particular patient, so that the right dose can be given the first time.