Allometric Scaling in Pharmacokinetics: Equation, Exponents, and Where It Fails

Allometric scaling is the standard method preclinical pharmacokinetics teams use to extrapolate a drug’s disposition parameters — most commonly clearance (CL) and, to a lesser extent, volume of distribution (Vd) — from animal species to humans, before any dose has ever been given to a person. It rests on a simple observation: many physiological and pharmacokinetic parameters scale with body size in a predictable, non-linear way across mammals. That predictability is real and useful, but it is also incomplete, and the pharmacokinetics literature has spent four decades documenting exactly where and why it breaks down, and building correction methods to compensate. This guide covers the equation, the conventional exponent values, the specific failure modes that are well established in the literature, and the refinements — correction factors, the rule of exponents, and multi-species approaches — that were developed in direct response to those failures.

The general allometric equation

Simple allometric scaling fits the power-function form:

Y = a × BW^b

where Y is the pharmacokinetic parameter of interest, BW is body weight, a is the allometric coefficient (the y-intercept on a log-log plot), and b is the allometric exponent (the slope). In practice, the parameter is measured in several species — typically mouse, rat, rabbit, dog, monkey, and sometimes additional species — and plotted against body weight on log-log axes. Linear regression on the log-transformed data yields log(a) and b, and the fitted line is then extrapolated to human body weight to generate a predicted human value.

This approach, and the broader observation that metabolic and physiological processes scale non-linearly with body mass, traces to Kleiber’s work on basal metabolic rate in the 1930s and was formalized for pharmacokinetics by Boxenbaum in the early 1980s (Boxenbaum H, J Pharmacokinet Biopharm 1982), who described it as reflecting a shared “physiological time” across species rather than a coincidental curve fit.

Typical exponent conventions

The exponent b is not arbitrary — different pharmacokinetic parameters have characteristic ranges, reflecting the underlying physiology being scaled:

• Clearance typically scales with b in the region of 0.6–0.8, echoing the ~0.75 exponent long observed for basal metabolic rate (Kleiber’s law) and for organ blood flows, since clearance is fundamentally limited by the perfusion and enzymatic capacity of eliminating organs.
• Volume of distribution scales closer to b ≈ 1.0, because distribution volume tracks anatomical space (extracellular water, tissue mass, plasma volume) more directly than it tracks a rate process, and correlates with body weight consistently across species.
• Half-life, because it is a hybrid, derived parameter (t½ is proportional to Vd/CL, not measured independently), correlates poorly with body weight on its own and is not itself a reliable scaling target — it is generally back-calculated from separately scaled CL and Vd rather than allometrically scaled directly.

Because Vd tends to scale close to linearly with body weight, it is the better-behaved of the two core parameters across species. Clearance is where simple allometry runs into trouble, and where the bulk of the methodological refinement described below has been directed.

Where simple allometry documented failure modes actually occur

Simple two-parameter allometry (BW and CL or Vd alone) is well documented in the pharmacokinetics literature to under- or over-predict human clearance by a wide margin for a substantial share of compounds — published reviews report prediction errors exceeding 30% relative to observed human clearance in a majority of cases when the basic equation is applied without correction. The reason is structural, not statistical noise: body weight alone does not capture cross-species differences in the actual biological machinery that eliminates a drug.

The specific mechanisms behind this gap are well characterized:

• Metabolic enzyme expression and activity differ across species in ways body weight cannot capture. Cytochrome P450 isoform abundance, substrate specificity, and catalytic activity vary substantially between rodents, dogs, non-human primates, and humans — two species of similar body weight can have very different intrinsic clearance for a compound cleared by a specific CYP pathway, because the enzyme itself behaves differently, not because of a size mismatch.
• Renally cleared drugs are a well-documented weak point for simple allometry, since active tubular secretion and reabsorption transporter expression and affinity differ across species independent of body size or glomerular filtration rate scaling.
• Plasma protein binding differences across species change the free (unbound) fraction available for elimination, which body-weight scaling does not account for on its own — this is part of why parameters are sometimes scaled on an unbound basis.
• Compounds eliminated primarily by phase I oxidative (hepatic) metabolism are specifically called out in the literature as a class where basic two-term allometry is unreliable enough that a dedicated correction method (the brain-weight-based two-term equation, below) was developed.

None of this means allometric scaling is unreliable in general — it means simple, uncorrected allometry is a first-pass estimate with a known, quantified error rate for certain elimination pathways, and the field’s response has been to build correction methods aimed specifically at those pathways rather than to abandon the underlying approach.

The rule of exponents and correction methods

The most widely cited response to simple allometry’s clearance-prediction problem is the rule of exponents, developed by Mahmood and Balian (Clin Pharmacokinet 1999; see also Mahmood, Adv Drug Deliv Rev 2007). Rather than treating every compound identically, the rule uses the value of the fitted exponent b itself as the signal for which correction, if any, a given compound needs before its clearance prediction is trusted: compounds whose exponent falls in a “well-behaved” mid-range are taken as adequately described by simple allometry, while compounds whose exponent falls outside that range are flagged for a correction method rather than accepted at face value. The specific correction methods that have been developed and validated against observed human data include:

• Maximum life-span potential (MLP) correction — sometimes described under the “neoteny” concept — predicts clearance using both species body weight and each species’ maximum recorded life-span, on the reasoning that life-span itself correlates with metabolic rate and elimination capacity in ways raw body weight alone does not.
• Brain weight correction — a two-term power equation incorporating both brain weight and body weight, used specifically to predict the intrinsic clearance of drugs eliminated primarily via phase I oxidative metabolism, the class identified above as a particular weak point for the basic equation.
• The product of clearance and brain weight as an alternative correction term, proposed as a further refinement for the same oxidative-metabolism compound class.
• Multiple-species allometry with in vitro data integration — incorporating in vitro metabolic stability or intrinsic clearance data (e.g., from hepatocytes or liver microsomes across the same species used for the in vivo allometry) alongside the body-weight regression, rather than relying on body weight as the sole predictor. This IVIVE-informed approach has generally improved predictive performance over body-weight-only scaling in published comparisons.

The common thread across all of these refinements is the same one driving the failure-mode list above: they each add a term or a data source that carries information about metabolic/physiological capacity that body weight, on its own, does not encode.

How this fits into preclinical-to-clinical development

Allometrically scaled clearance and volume of distribution feed into first-in-human dose projections alongside — not instead of — toxicology-derived safety margins; see NOAEL and first-in-human starting dose for how the two streams combine. Allometric scaling itself sits within the broader discipline of toxicokinetics in nonclinical safety studies, which generates the multi-species exposure data the allometric regression is built on, and within the wider field of pharmacometrics, where allometric relationships are also used as covariate models within population PK. For compounds where mechanistic understanding of species differences in metabolism and transport matters more than an empirical body-weight fit can capture, physiologically based pharmacokinetic (PBPK) modeling is increasingly used as a complement or alternative — it explicitly models the organ-level physiology and enzyme kinetics that simple allometry can only approximate through the body-weight term. Converting between plasma- and blood-referenced clearance values before running the regression also matters; see blood-to-plasma ratio conversions in PK for that step. Readers newer to the underlying discipline may want the grounding in what pharmacology covers as a field before working through species-scaling methodology specifically.

Frequently asked questions

Why does volume of distribution scale better across species than clearance does?

Volume of distribution reflects anatomical and physicochemical space — extracellular water, tissue mass, plasma volume, and how the compound partitions into those compartments — all of which correlate closely with body size across mammals. Clearance instead depends on the biological activity of specific eliminating organs (their enzyme content, transporter expression, blood flow), which does not track body size as tightly. That is why clearance, not volume of distribution, is where the correction methods above concentrate.

Does a poor allometric fit mean the drug can’t be scaled at all?

No — it means uncorrected two-parameter allometry (body weight alone) is not reliable for that compound, not that cross-species extrapolation is impossible. The rule of exponents exists specifically to flag which compounds need a correction method (MLP, brain weight, or in vitro data integration) rather than being taken at face value from the raw body-weight regression, and PBPK modeling is available as a further, more mechanistic alternative when even corrected allometry is judged insufficient.

How many species are typically used in the allometric regression?

Most published applications use at least three, and commonly four to six, species spanning a wide body-weight range (for example mouse, rat, rabbit, dog, and monkey) to establish a stable log-log regression before extrapolating to human body weight. A regression built on only two species is generally considered too sensitive to species-specific noise to trust for a human prediction.

Is allometric scaling still relevant now that PBPK modeling is available?

Yes. Allometric scaling remains widely used because it requires far less mechanistic data and is faster to apply than a full PBPK model, making it a standard first-pass tool early in preclinical development. PBPK modeling is generally reserved for compounds where mechanistic questions (specific enzyme or transporter involvement, nonlinear kinetics, drug-drug interaction risk) make an empirical body-weight fit insufficient on its own.

Primary sources: Boxenbaum H, “Interspecies scaling, allometry, physiological time, and the ground plan of pharmacokinetics,” Journal of Pharmacokinetics and Biopharmaceutics (1982); Mahmood I, Balian JD, “The pharmacokinetic principles behind scaling from preclinical results to phase I protocols,” Clinical Pharmacokinetics (1999); Mahmood I, “Application of allometric principles for the prediction of pharmacokinetics in human and veterinary drug development,” Advanced Drug Delivery Reviews (2007).