Technical Bulletin · Human PK Translation

From Preclinical ADME to Human PK: Choosing Between IVIVE, Allometry, and PBPK

A question-first framework for matching available molecular, in vitro, animal and human data to an appropriate method for predicting clearance, distribution and exposure.

Discuss Your Human PK Strategy

Opening answer

Choose the method from the question and the evidence—not from the tool name

No single translation method is best for every compound or development stage. Structure-based QSAR or machine learning is useful when chemistry is available but experimental data are sparse. IVIVE becomes more informative when human-relevant in vitro clearance, binding, permeability or transporter measurements are available. Allometry uses animal in vivo PK to scale parameters across body size, whereas PBPK integrates drug-specific evidence with physiology to simulate concentration–time behavior and test mechanistic scenarios. Population PK begins after human concentration data exist and estimates typical parameters, variability and covariate effects in the studied population. The practical rule is available data + development question → appropriate approach. Predicting a rough clearance range for compound selection, proposing a first-in-human exposure profile, explaining nonlinear absorption, and individualizing dose in patients are different tasks. Each requires different inputs, assumptions, verification and tolerance for uncertainty. Methods are often combined, but their outputs should not be treated as independent confirmation when they rely on the same underlying data.

Start with the decision

Define what must be predicted before selecting the method

Four framing questions

A method is appropriate only if its inputs can identify the processes that control the requested output.

What data actually exist?

Structure only, human in vitro ADME, animal PK, formulation data, or observed human concentrations?

What output is needed?

A rank, CL or Vd estimate, oral exposure, full concentration–time curve, or population variability?

What decision follows?

Compound prioritization, candidate selection, first-in-human planning, scenario evaluation or dose adjustment?

Also define route, dose range, formulation, target population and acceptable error. A model that supports relative ranking among analogues may be unsuitable for an absolute human dose projection. Conversely, a mechanistic PBPK model can be unnecessary when the decision only requires separating low- and high-clearance compounds in an early series.

Method landscape

Five approaches occupy different positions in the evidence chain

Structure-led

QSAR / ML

Learns relationships between molecular representation and measured properties. Best for early ranking or filling data gaps within a defined applicability domain.

Human in vitro

IVIVE

Scales measured intrinsic processes, such as hepatic metabolism, to an expected human organ or whole-body parameter using physiological factors and a disposition model.

Animal in vivo

Allometry

Relates PK parameters across species to body weight or another size descriptor. It is empirical and assumes the cross-species relationship remains informative in humans.

Mechanistic integration

PBPK

Combines physiology, physicochemical properties and ADME mechanisms to simulate concentration–time profiles across doses, routes, tissues or populations.

Human observations

PopPK

Fits human concentration data to estimate typical PK, interindividual variability, residual variability and covariate relationships.

These labels do not represent a simple competition. QSAR predictions can provide provisional PBPK inputs; IVIVE can supply hepatic or renal clearance; allometry may offer a comparator for CL or Vd; and early human data can update a PBPK model or seed a PopPK analysis. The value lies in preserving which evidence is measured, which is scaled and which is assumed.

PK quantities

Parameters and exposure summaries are related, but they are not interchangeable

Core quantities

Interpret each value with route, dose, units, model structure and sampling context.

CLClearance: proportionality between elimination rate and concentration.

VdApparent volume linking amount in the body to measured concentration.

FBioavailability: fraction of an administered extravascular dose reaching systemic circulation.

t½Half-life: time-related decline governed by both distribution and clearance; for a simple linear one-compartment case, t½ ≈ 0.693 × V/CL.

CmaxObserved or predicted peak concentration, influenced by dose, absorption, F, distribution and sampling.

AUCTotal systemic exposure over a defined interval; under linear conditions, dose, F and CL are primary determinants.

Conceptual illustration only. AUC does not uniquely define peak, trough or time above a pharmacological threshold.

A full concentration–time profile is needed when peak-related safety, trough coverage, accumulation, dosing interval or absorption rate matters. A single CL estimate cannot define oral Cmax without assumptions or measurements for absorption, F, distribution and dose. Likewise, a reported terminal half-life can reflect multicompartment distribution and study design rather than a single Vd/CL relationship.

Decision component

Human PK method-selection table

Use the table as a starting point, then test whether the required input assumptions are identifiable for the compound and intended decision.

Approach	Minimum useful input	Best-matched question	Typical outputs	Main assumptions and limits	Evidence escalation
QSAR / ML	Standardized structure plus a model trained on a relevant endpoint and represented chemistry	Which compounds are likely to have more favorable PK-related properties?	Property class or estimate; sometimes CL, Vd, F or half-life proxies	Model output is not a measurement; depends on chemical-space coverage, endpoint definition and validation design. Full profiles require additional assumptions.	Measure the influential ADME properties and update ranking or downstream models.
IVIVE	Human microsomal/hepatocyte or other system-specific intrinsic data, binding and physiological scaling factors	What human clearance or pathway contribution is suggested by measured in vitro activity?	Intrinsic, hepatic, renal or total CL estimates; sometimes extraction ratio	Requires system correction, recovery, binding treatment and a suitable liver/organ model. Extrahepatic or transporter effects may be missed.	Compare across systems, examine empirical scaling factors cautiously, and benchmark against animal or emerging human PK.
Allometry	Reliable IV or oral PK from multiple informative species with comparable parameter definitions	How might CL or Vd scale from animal species to humans?	Human CL, Vd and derived half-life range	Empirical body-size relationship; vulnerable to species-specific metabolism, active transport, protein binding, nonlinear PK and oral absorption differences.	Investigate discordant species and compare with IVIVE or mechanism-based estimates.
PBPK	Physicochemical, formulation, binding, permeability/absorption, distribution and clearance evidence appropriate to the question	What concentration–time profiles or untested mechanistic scenarios are plausible?	Plasma/tissue profiles, Cmax, AUC, half-life, accumulation and scenario comparisons	Data- and assumption-intensive; uncertain parameters may be non-identifiable. Verification must challenge the mechanisms supporting the intended use.	Verify the base model with independent PK and perform sensitivity and uncertainty analyses before extrapolation.
Population PK	Human concentration–time observations, dosing records and relevant covariates	What are typical PK, variability and covariate effects in the observed population?	Population CL/V parameters, variability, covariate relationships and individual predictions	Cannot replace pre-human translation because it requires human data. Sparse sampling and confounded covariates can limit identifiability.	Use prospective sampling, external evaluation and exposure–response analysis as development progresses.

IVIVE or allometry?

The source of biological information determines the failure mode

Prefer IVIVE when

Human-relevant in vitro systems capture the major metabolic or transport processes and the question centers on clearance mechanism. IVIVE avoids assuming that an animal's pathway mix matches the human pathway mix, but it remains sensitive to assay recovery, binding, enzyme abundance, scaling and model choice.

Prefer allometry when

Comparable systemic PK exists in more than one informative species and parameters scale consistently. Allometry can be practical for Vd and CL ranges, but a visually good weight relationship does not prove conserved biology. Species-specific enzymes or transporters can create confident-looking but biased extrapolation.

Use both as triangulation when

The methods contain meaningfully different evidence. Agreement can increase confidence only after shared inputs are identified; disagreement is diagnostically useful. It may reveal underprediction in an in vitro system, species-specific clearance, unmodeled renal or biliary elimination, or inconsistent parameter estimation.

When PBPK adds value

Escalate from parameter scaling to profile simulation for a defined reason

Absorption is mechanistic

Solubility, dissolution, permeability, gut metabolism, formulation or food effects may shape oral exposure and Cmax.

Processes interact

Metabolism, transport, binding, tissue distribution or nonlinear kinetics cannot be interpreted independently.

Scenario matters

Dose, route, schedule, age, organ impairment or another population changes the development question.

The curve is the output

Peak, trough, accumulation, tissue exposure or time above a threshold matters more than one average parameter.

PBPK is not automatically superior when more data become available. A model with many weakly informed inputs may be less decision-useful than a transparent IVIVE or allometric range. Before adding complexity, define the intended use and identify which simulated mechanisms can be verified. Parameter optimization should be biologically justified and documented; fitting several uncertain inputs to one observed curve can produce non-unique solutions.

PopPK is the next evidence layer, not a preclinical alternative. Once human data exist, PopPK can quantify between-subject variability and covariates. PBPK and PopPK may then be complementary: one encodes mechanistic prior knowledge and untested scenarios, while the other estimates patterns supported by observed human concentrations.

Input quality and verification

What must be documented before a human PK projection is used

Compound and formulation identity. Salt, stereochemistry, solid form, dose form and concentration conditions match the modeled question.
Route-specific evidence. IV and oral data are not treated as interchangeable; F and absorption assumptions are explicit.
Assay provenance. Species, matrix, protein/cell concentration, binding, recovery, units and transformation are traceable.
Clearance accounting. Hepatic, renal, biliary and other pathways are considered rather than forcing all loss into one mechanism.
Parameter definitions. CL, Vd and half-life estimates use comparable models and sampling windows across datasets.
Applicability and scaling. QSAR domain, IVIVE scaling factors and allometric species selection are justified.
Sensitivity and uncertainty. Influential inputs are varied over plausible ranges and propagated to decision-relevant outputs.
Independent evaluation. Data used to adjust the model are distinguished from data used to test its predictive behavior.
Decision criteria. Required precision and consequences of under- or overprediction are stated before interpreting results.
Versioned reporting. Data sources, equations, software/model version, assumptions and simulations are reproducible.

Practical selection examples

Match common development situations to the least complex adequate approach

Available evidence	Immediate question	Reasonable starting approach	What would change the choice?
Structures for hundreds of analogues; limited measured ADME	Which compounds should enter experimental profiling?	Domain-aware QSAR/ML ranking with diversity sampling	New assays enable local model updates and IVIVE for measured compounds.
Human hepatocyte stability, plasma binding and basic physicochemical data	What is a plausible human hepatic CL range?	IVIVE with alternative well-stirred assumptions and sensitivity analysis	Evidence of renal, transporter-mediated or extrahepatic clearance may require additional experiments or PBPK.
Consistent IV PK in three species	What CL and Vd range might support first-dose planning?	Allometry, compared with IVIVE where possible	Species-specific metabolism or nonlinear PK reduces confidence and favors mechanistic investigation.
Rich ADME data plus oral absorption and animal PK	What human profile is plausible across doses or formulations?	Fit-for-purpose PBPK with prespecified verification	If only a simple absolute parameter is needed, IVIVE/allometry may remain adequate.
Early clinical concentrations with heterogeneous dosing and sampling	What explains variability, and which covariates affect exposure?	Population PK, potentially informed by prior mechanistic knowledge	Mechanistic extrapolation to unobserved physiology may justify a linked PBPK analysis.

Relevant services

Connect the available evidence to an appropriate human PK analysis

A useful project starts by auditing the available structures, in vitro measurements, animal PK and human observations against the decision to be supported. These existing services align with early parameter prediction and mechanistic profile simulation.

Pharmacokinetic (PK) Prediction

PK-oriented prediction and integration of molecular, ADME and preclinical evidence for compound comparison and human translation planning.

Explore the service →

PBPK Modeling and Simulation

Fit-for-purpose mechanistic modeling for concentration–time profiles, sensitivity analysis and scenario evaluation.

Explore the service →

Review Your Available PK Data

References

Scientific and Regulatory Sources

U.S. Food and Drug Administration. Physiologically Based Pharmacokinetic Analyses—Format and Content: Guidance for Industry. September 2018.
U.S. Food and Drug Administration. Population Pharmacokinetics: Guidance for Industry. February 2022.
European Medicines Agency. Guideline on the Reporting of Physiologically Based Pharmacokinetic (PBPK) Modelling and Simulation. 2018.
Yim, D.-S. Predicting human pharmacokinetics from preclinical data. Translational and Clinical Pharmacology 2021, 29, 54–68. doi:10.12793/tcp.2021.29.e11.
Obach, R. S.; Baxter, J. G.; Liston, T. E.; et al. The prediction of human pharmacokinetic parameters from preclinical and in vitro metabolism data. Journal of Pharmacology and Experimental Therapeutics 1997, 283, 46–58.
Hosea, N. A.; Collard, W. T.; Cole, S.; et al. Prediction of human pharmacokinetics from preclinical information: comparative accuracy of quantitative prediction approaches. Journal of Clinical Pharmacology 2009, 49, 513–533. doi:10.1177/0091270009333209.
Ito, K.; Houston, J. B. Prediction of human drug clearance from in vitro and preclinical data using physiologically based and empirical approaches. Pharmaceutical Research 2005, 22, 103–112. doi:10.1007/s11095-004-9009-0.
Jones, H. M.; Chen, Y.; Gibson, C.; et al. Physiologically based pharmacokinetic modeling in drug discovery and development: a pharmaceutical industry perspective. Clinical Pharmacology & Therapeutics 2015, 97, 247–262. doi:10.1002/cpt.37.

Online Inquiry

Submit your project details below, and our team will respond within 24 hours.

Free

Project Feasibility Assessment

Free

Project Sample

Need help getting the data you need?

Talk to our technical team about your project!

I Want To Talk