Population Pharmacokinetics
Population Pharmacokinetics (Nonlinear Mixed-Effects) · Also known as: PopPK, Nonlinear Mixed-Effects Modeling, NONMEM Approach, Popülasyon Farmakokinetiği
Population Pharmacokinetics (PopPK) is a nonlinear mixed-effects modeling framework that characterizes how drugs are absorbed, distributed, metabolized, and eliminated across a patient population, estimating both typical population parameters and the magnitude of between-subject variability. Introduced by Sheiner, Rosenberg, and Marathe in 1977, it enables parameter estimation from sparse, routinely collected clinical data—making it indispensable in drug development, regulatory submissions, and individualized dosing.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use PopPK when drug concentration data are sparse or unbalanced across individuals, as is typical in clinical trials, pediatric studies, or routine therapeutic drug monitoring. The method requires a plausible structural PK model, an assumption that between-subject variability follows a log-normal or normal distribution, and that residual errors are independent. It is inappropriate when sample sizes are very small (fewer than ~20 subjects) without informative priors, or when the structural model is fundamentally misspecified. Alternatives for rich, balanced designs include naive pooled or two-stage approaches.
Strengths & limitations
- Efficiently uses sparse, unbalanced concentration-time data that would defeat individual-level fitting
- Simultaneously estimates population typical values, between-subject variability, and covariate effects in one inferential step
- Supports Bayesian individual parameter estimation (empirical Bayes), enabling model-informed precision dosing
- Provides a rigorous statistical framework accepted by FDA and EMA for regulatory pharmacometric submissions
- Requires a correctly specified structural pharmacokinetic model; misspecification propagates into all parameter estimates
- Computational cost is high for complex models with many random effects or large datasets; convergence is not guaranteed
- Covariate selection can be data-driven and prone to overfitting without pre-specified hypotheses or penalization
- Interpretation of omega (between-subject variance) and sigma (residual variance) requires careful model diagnostics and expert judgment
Frequently asked
How many subjects and samples are needed for a reliable PopPK analysis?
As a general guideline, at least 50–100 subjects with two or more observations per subject provide stable estimates of typical parameters and between-subject variability. Fewer subjects are feasible when informative prior distributions are incorporated in a Bayesian framework, but very small datasets risk poorly identified variance parameters and unreliable covariate effects regardless of the estimation algorithm used.
What is the difference between FOCE and SAEM estimation?
FOCE (First-Order Conditional Estimation) linearizes the nonlinear model around conditional individual parameter estimates, which is fast but can be biased for highly nonlinear models or large random effects. SAEM (Stochastic Approximation EM) uses a stochastic simulation approach to integrate over the random-effect distribution exactly, offering better convergence properties for complex models at higher computational cost.
Can PopPK handle non-normal between-subject variability?
Standard implementations assume log-normally distributed PK parameters (equivalent to normally distributed log-transformed parameters), which is appropriate for strictly positive quantities like clearance and volume. Extensions using mixture models or heavy-tailed distributions can accommodate bimodal or outlier-prone populations, but require larger datasets and careful identifiability assessment to avoid overfitting the variance structure.
Sources
- Sheiner, L. B., Rosenberg, B., & Marathe, V. V. (1977). Estimation of population characteristics of pharmacokinetic parameters from routine clinical data. Journal of Pharmacokinetics and Biopharmaceutics, 5(5), 445–479. DOI: 10.1007/BF01061728 ↗
How to cite this page
ScholarGate. (2026, June 2). Population Pharmacokinetics (Nonlinear Mixed-Effects). ScholarGate. https://scholargate.app/en/pharmacometrics/population-pharmacokinetics
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Bayesian Hierarchical ModelBayesian↔ compare
- Pharmacokinetic Compartment ModelPharmacometrics↔ compare