PBPK modelling

Every model covered elsewhere in this pillar treats the body as abstract boxes — a "central compartment," a "peripheral compartment" — that a drug flows between according to fitted equations. It works, and it predicts clinical data well, but it's abstract: you can't point to the peripheral compartment on an anatomy diagram.

Physiologically based pharmacokinetic (PBPK) modelling does something different. You build a model that maps directly onto real anatomy — the lungs, liver, kidneys, gut, muscle, fat, brain, heart — each as its own compartment, carrying real measured physiology: blood flow to that organ, enzyme activity there, plasma protein binding, transporter expression at the gut wall. Then you put a candidate drug into this virtual body, described only by its own physical and chemical properties — molecular weight, lipophilicity, protein binding — and simulate where it goes and how it's cleared, without having run a single clinical study yet.

That's what makes PBPK unusually useful: it can predict drug behaviour in situations you have no data for at all — a paediatric patient, someone with kidney failure, a different ethnic population, a second drug on board — all from first principles.

Where it sits in development#

PBPK is the one modelling approach that genuinely spans the whole pipeline: predicting human PK from lab data before any human trial in discovery, translating animal data into a first-in-human dose, predicting PK in special populations in Phase I to avoid running a dedicated study for each one, and supporting or replacing clinical drug-drug interaction studies in later development. The regulatory significance here is real — the FDA now explicitly accepts PBPK model predictions in place of certain clinical studies, which is a mathematical model substituting for a trial, not just informing one.

What's actually settled, and what isn't#

The core PBPK theory — perfusion-limited and permeability-limited organ models — dates to the 1970s, and small-molecule PBPK for adults is commercially mature, implemented in established software (Simcyp, GastroPlus, the open-source PK-Sim). What's still frontier:

• Pediatric PBPK has to model organ growth, enzyme maturation, and changing body composition as a child develops — current models approximate this, with the biggest gaps in neonates and infants.
• Biologic PBPK — modelling FcRn-mediated recycling and target-mediated tissue distribution for antibodies — is still evolving; the mechanisms differ enough from small molecules that the standard PBPK toolkit doesn't transfer cleanly.
• mRNA and gene therapy PBPK is close to undeveloped: how a lipid nanoparticle distributes, crosses into a cell, escapes the endosome, and behaves intracellularly doesn't have a validated modelling framework yet.
• Tumour PBPK — how a drug actually distributes inside a solid tumour, given uneven vasculature, elevated internal pressure, and necrotic regions — is an active oncology-adjacent frontier.
• Drug-disease interactions: liver cirrhosis, heart failure, obesity, and inflammation all change the physiological parameters PBPK models normally treat as fixed, and most drugs are labelled from healthy-adult data even though the patients who need them often aren't healthy adults.

Software licence costs and specific market-size figures for any of this change often enough that I'm not going to quote one here without a source to point at.