Pharmacometrics

Pharmacometrics is the science of building mathematical models of how drugs behave in the body and how bodies respond to them, and then using those models to make decisions about drug development and treatment.

15 frameworks, from 1930 to today · Pharmacology · Medicine & Health

still a working view ended · each leads to its entry below

Overview

Pharmacometrics is the science of building mathematical models of how drugs behave in the body and how bodies respond to them, and then using those models to make decisions about drug development and treatment1. Its practitioners are not primarily bench scientists; they are modelers who work with data from clinical trials, patient populations, and laboratory experiments to answer questions that would otherwise require enormous, expensive studies: What dose should be given to a child? How should a drug be adjusted for a patient with kidney failure2? Is a failed trial a failure of the drug or of the dose chosen? The field matters because most drugs fail in development, and many that succeed are used at doses chosen by habit rather than by quantitative reasoning. Pharmacometrics is the discipline that tries to replace habit with calculation.

The Roots: Compartmental Thinking and the Mathematics of Elimination

The earliest strand of pharmacometrics is pharmacokinetics—the study of what the body does to a drug. In the 1920s and 1930s, pharmacologists began describing drug concentrations in blood and tissues using simple mathematical models. The key idea was the compartment: a simplified "space" within the body that a drug enters and leaves at measurable rates. A one-compartment model treats the whole body as a single well-mixed volume; a two-compartment model adds a peripheral tissue compartment. These models allowed scientists to estimate clearance (how quickly the body removes a drug) and volume of distribution (how widely a drug spreads), parameters that could be compared across drugs and species.

A parallel development was the recognition that drug elimination is not always proportional to concentration. In 1913, Leonor Michaelis and Maud Menten had described enzyme kinetics in which the reaction rate saturates at high substrate concentrations. When pharmacologists applied this to drug metabolism—notably in the 1960s and 1970s with drugs like phenytoin and ethanol—they found that Michaelis–Menten kinetics explained why some drugs accumulate disproportionately when doses are increased. The mathematics was borrowed from enzymology, but the clinical implication was new: for a drug with saturable metabolism, a small dose increase can produce a large, dangerous rise in blood concentration. This insight became a standard part of dose selection.

Noncompartmental Analysis and the Rise of Clinical Pharmacokinetics

Not every pharmacokinetic question required a structural model. In the 1970s and 1980s, noncompartmental analysis (NCA) became a standard tool. NCA estimates exposure measures such as the area under the concentration–time curve (AUC) and maximum concentration (Cmax) directly from observed data using numerical integration, without assuming a compartmental structure. It is robust and widely used for regulatory submissions, but it describes rather than explains: it cannot predict what will happen at a dose not studied, or in a population not enrolled.

The limitation of NCA—and of fitting models to one patient at a time—was the problem that nonlinear mixed-effects modeling (NLME) solved. In the 1970s and 1980s, statisticians and pharmacologists developed methods to fit pharmacokinetic models simultaneously to data from many individuals, estimating both typical population parameters and the variability between patients. The approach, associated with software such as NONMEM (developed by Lewis Sheiner and Stuart Beal at the University of California, San Francisco, beginning in the late 1970s), treated each patient's parameters as a random deviation from a population mean. This made it possible to identify which patient characteristics—weight, age, kidney function, genotype—explained variability, and to quantify how much variability remained unexplained. Sheiner, a physician and engineer, was a central figure in articulating the philosophy behind this: that drug development should be a learning process, with models updated as data accumulate.

Pharmacodynamics: Modeling What the Drug Does

Pharmacokinetics describes concentration; pharmacodynamic modeling describes the relationship between concentration and effect123. Early pharmacodynamic models used simple equations—the Hill equation, borrowed from receptor theory, relates drug concentration to fractional occupancy or response. But clinical effects are often delayed relative to plasma concentration, because the drug must reach a site of action, or because the effect is mediated by a downstream cascade. In the 1980s and 1990s, modelers developed effect-compartment models and indirect-response models to capture these delays. A concrete example: the anticoagulant effect of warfarin, measured by INR, lags behind plasma warfarin concentration by hours to days; an indirect-response model that accounts for the turnover of clotting factors explains the lag and allows prediction of the time course of effect after a dose change.

The integration of pharmacokinetics and pharmacodynamics into a single PK/PD model became the core of pharmacometric reasoning: the model predicts the time course of effect, not just of concentration, and therefore can be used to simulate alternative dosing regimens.

Physiologically Based Pharmacokinetic Modeling

A different modeling tradition asked not "how many compartments?" but "what are the actual organs?" Physiologically based pharmacokinetic (PBPK) modeling represents the body as a network of anatomically defined compartments—liver, kidney, muscle, fat, brain—connected by blood flow, each with its own volume and partition coefficient. The approach has roots in the 1930s work of Torsten Teorell, who proposed a circulation-based model of drug distribution, but it matured in the 1990s and 2000s as computing power and physiological data improved. PBPK's distinctive strength is extrapolation: because its parameters have physiological meaning, it can predict drug behavior in populations not studied—children, pregnant women, patients with organ impairment—by adjusting organ sizes, blood flows, and enzyme activities. Regulatory agencies now accept PBPK analyses in lieu of some clinical studies, particularly for pediatric dosing and drug–drug interaction assessment.

Bayesian Thinking and Optimal Design

A parallel thread in pharmacometrics is Bayesian statistics—the framework in which prior knowledge is combined with new data to produce updated probability distributions for parameters4. Bayesian methods entered pharmacokinetics in the 1970s and 1980s, notably through the work of Sheiner and colleagues on Bayesian adaptive control of drug dosing: individualizing a patient's dose by combining a population model (the prior) with that patient's measured drug concentrations (the likelihood). This was applied clinically in therapeutic drug monitoring for drugs with narrow therapeutic windows, such as aminoglycosides and immunosuppressants.

Optimal design theory, developed in the statistical literature and adapted to pharmacometrics, asked a related question: given a model and a set of candidate sampling times, which times yield the most information about the parameters? This allowed trial designers to reduce the number of blood samples per patient while preserving the ability to estimate key parameters—a practical advance that made population pharmacokinetic studies feasible in settings where intensive sampling was impossible.

Model-Informed Drug Development and Meta-Analysis

By the 2000s, the pieces had come together into a framework now called model-informed drug development (MIDD). The idea is that models should inform decisions at every stage: dose selection for first-in-human trials, trial simulation to choose sample sizes and endpoints, bridging strategies to support approval in new populations, and label optimization. The U.S. Food and Drug Administration and the European Medicines Agency began formal programs to review pharmacometric analyses as part of regulatory submissions; the FDA's MIDD pilot program, launched in 2018, explicitly invited sponsors to use models to support dosing decisions.

A related development was model-based meta-analysis, which fits a single model to summary-level data from many published trials, allowing comparison of drugs and doses across studies that were never designed to be compared. This approach became influential in therapeutic areas where head-to-head trials are rare, such as diabetes and rheumatoid arthritis, by quantifying dose–response relationships across competing agents.

The Present: Simulation, Systems, and the Limits of Prediction

Contemporary pharmacometrics is dominated by simulation. A typical analysis builds a population PK/PD model, then runs thousands of virtual trials under different dosing scenarios to estimate the probability of achieving a target exposure or response. This "clinical trial simulation" approach is now standard in dose selection and is increasingly used to support labeling decisions for special populations.

The field's open questions are substantial. How much can models predict before data exist—for a new drug in a new population, or for a drug with a novel mechanism? PBPK and quantitative systems pharmacology (which adds mechanistic biological networks to the pharmacokinetic framework) push toward greater mechanistic detail, but each added parameter demands data that may not be available. The tension between parsimony and realism is unresolved. A second open question is the integration of pharmacogenomic data: models can include genotype as a covariate, but the effect sizes for most gene–drug pairs remain uncertain, and the clinical utility of genotype-guided dosing is established for only a minority of drugs. A third is the validation of models themselves: how does one know that a model's predictions are trustworthy? The field has developed diagnostic tools—visual predictive checks, bootstrap confidence intervals—but no universal standard for model credibility exists, and regulatory acceptance still depends on case-by-case review.

The through-line from the 1920s compartment to today's virtual patient is a single ambition: to make the behavior of drugs in the body calculable, and therefore predictable, before the patient is exposed.

References

  1. Pharmacometrics (introduction, Pharmacodynamics (PD)), Wikipedia
  2. Pharmacokinetics (Population pharmacokinetics, introduction), Wikipedia
  3. Pharmacodynamics, Wikipedia
  4. Bayesian statistics, Wikipedia

The frameworks of Pharmacometrics

Each framework is a way of working the field has used: when it held, and the question it answers.

  1. 1930-Present

    Compartmental pharmacokinetics

    How does drug concentration change over time?

    Fly to its tower in Nousfarer

  2. 1960-Present

    Pharmacodynamic modeling

    How does drug concentration relate to effect?

    Fly to its tower in Nousfarer

  3. 1965-Present

    Michaelis–Menten kinetics

    Why does elimination saturate at high concentrations?

    Fly to its tower in Nousfarer

  4. 1970-Present

    Noncompartmental analysis

    How much drug exposure occurred without a model?

    Fly to its tower in Nousfarer

  5. 1975-Present
  6. 1977-Present
  7. 1980-Present
  8. 1980-Present

    PKPD model

    What effect will a dosing regimen produce?

    Fly to its tower in Nousfarer

  9. 1989-Present

    Learn-confirm paradigm

    How should drug development learn and confirm?

    Fly to its tower in Nousfarer

  10. 1990-Present
  11. 1995-Present

    Bayesian pharmacometrics

    How should prior knowledge update with data?

    Fly to its tower in Nousfarer

  12. 1995-Present
  13. 2000-Present

    Model-based meta-analysis

    How do drugs compare across many trials?

    Fly to its tower in Nousfarer

  14. 2000-Present

    Model-informed drug development

    How should models guide drug development?

    Fly to its tower in Nousfarer

  15. 2010-Present

    Quantitative systems pharmacology

    How does drug action unfold mechanistically?

    Fly to its tower in Nousfarer

How the ideas connect

Each arrow runs from an earlier idea to the later one it shaped. The list below says what passed between them.

Compartmental pharmacokineticsPharmacodynamic modelingMichaelis–Menten kineticsNoncompartmental analysisModel-Informed Precision DosingNonlinear mixed-effects modelingOptimal experimental designPKPD modelLearn-confirm paradigmPhysiologically based pharmacokinetic…Bayesian pharmacometricsClinical trial simulationModel-based meta-analysisModel-informed drug developmentQuantitative systems pharmacology

Beyond the field

Where Pharmacometrics's ideas came from in other fields, and what they led to there: the strongest 12 of 90 links.

A city in Nousfarer: a red-carpeted stair climbing terraces to a crowned tower on an acropolis, more towers round it, forest and snowy mountains beyond.
A city in Nousfarer, the game played in the browser.

Pharmacometrics is a planet in Nousfarer

Its city has a tower for each of these 15 frameworks. Land, walk in, and climb one concept at a time.

Land on Pharmacometrics

Related fields

Fields its ideas are linked to

More in Pharmacology

See where Pharmacometrics sits on the Map of Human Knowledge