Choosing a Body-Size Descriptor for Allometry

Why the Size Descriptor Matters

An allometric model relates a pharmacokinetic parameter to a measure of body size. Choosing that measure is not a cosmetic decision: two descriptors can rank people differently or imply different degrees of size change. As a result, they can lead to different predicted clearances or volumes, even when the allometric exponent is held fixed.

Total body weight is easy to measure and includes all body mass. Body surface area (BSA) combines weight and height into a geometric measure. Fat-free mass (FFM) estimates mass excluding fat and can distinguish people whose total weights are similar but whose body composition differs. None is automatically the correct descriptor for every drug, population, or pharmacokinetic parameter. The choice should reflect the scientific setting and be checked against data relevant to the intended use.

Definition: A body-size descriptor is a measured or calculated quantity used to represent body size in a model. In a power-law allometric model, a parameter \(P\) is related to descriptor \(S\) by \(P=kS^\theta\), where \(k\) is a constant and \(\theta\) is the allometric exponent. The descriptor \(S\) may be total body weight, BSA, or FFM; its definition and units must be stated.

For a transparent comparison, suppose alternative models use the same exponent and are calibrated to give the same parameter value for a chosen individual A. The predicted ratio for another individual B is then \((S_B/S_A)^\theta\). This ratio isolates the consequence of descriptor choice. It does not establish which descriptor is biologically best, and it is not a substitute for evaluating a model against data.

Defining the Descriptors

Total body weight, denoted \(W\), is recorded in kilograms. It is direct and widely available, but it does not distinguish fat mass from lean tissue or account explicitly for height. BSA is often calculated from height and weight. One common adult formula is the Mosteller equation, with height in centimetres and weight in kilograms:

BSA=\sqrt{\frac{H_{\mathrm{cm}}W_{\mathrm{kg}}}{3600}}\quad \mathrm{m^2}.

FFM is not generally measured directly in routine pharmacometric data sets. It is often estimated from weight, height, and a body-composition equation. For illustration, the adult Janmahasatian equations are:

FFM_{\mathrm{male}}= \frac{9270W}{6680+216\,BMI}, \qquad FFM_{\mathrm{female}}= \frac{9270W}{8780+244\,BMI}, \qquad BMI=\frac{W}{H_{\mathrm{m}}^2}.

Here weight is in kilograms, height is in metres, BMI is in \(\mathrm{kg/m^2}\), and the resulting FFM is in kilograms. These are estimation equations, not direct measurements or universal definitions of body composition. Their use should be appropriate to the population and available data.

Model comparison condition: When comparing descriptors through predicted ratios, specify the exponent, the parameter being predicted, and the individual used as the common anchor. If \(P=kS^\theta\) and the model gives \(P_A\) for individual A, then \(P_B=P_A(S_B/S_A)^\theta\). The comparison assumes the same exponent and anchor parameter across the alternatives; changing either at the same time would mix effects.

Worked Comparisons

Worked Example: Clearance and Volume Under Three Descriptors

Consider two hypothetical adult men. Individual A weighs 70 kg and is 175 cm tall; individual B weighs 100 kg and is 180 cm tall. For this comparison, take the clearance exponent as 0.75 and the volume exponent as 1.0. These are assumed values for illustration, not estimates from the two individuals. Suppose A has clearance 5.00 L/h and volume 50.0 L under each candidate model.

First calculate BSA. For A, \(BSA_A=\sqrt{175(70)/3600}=\sqrt{3.40278}=1.84466\ \mathrm{m^2}\). For B, \(BSA_B=\sqrt{180(100)/3600}=\sqrt{5}=2.23607\ \mathrm{m^2}\). Thus \(BSA_B/BSA_A=2.23607/1.84466=1.2122\). The displayed operands and ratio use the same rounded values.

For FFM, A has \(BMI_A=70/1.75^2=22.8571\ \mathrm{kg/m^2}\), so \(FFM_A=9270(70)/[6680+216(22.8571)]=648900/11617.1=55.86\ \mathrm{kg}\). B has \(BMI_B=100/1.80^2=30.8642\ \mathrm{kg/m^2}\), so \(FFM_B=9270(100)/[6680+216(30.8642)]=927000/13346.7=69.46\ \mathrm{kg}\). Using the displayed rounded FFM values, \(FFM_B/FFM_A=69.46/55.86=1.243\). The weight ratio is \(100/70=1.429\).

For clearance, raise each descriptor ratio to 0.75. The weight-based prediction is \(5.00(1.429)^{0.75}=6.53\ \mathrm{L/h}\); the BSA-based prediction is \(5.00(1.2122)^{0.75}=5.78\ \mathrm{L/h}\); and the FFM-based prediction is \(5.00(1.243)^{0.75}=5.89\ \mathrm{L/h}\). For volume, the exponent is 1.0, so the predictions are \(50.0(1.429)=71.4\ \mathrm{L}\), \(50.0(1.2122)=60.6\ \mathrm{L}\), and \(50.0(1.243)=62.2\ \mathrm{L}\), respectively.

The results differ because these descriptors represent the size change differently. Weight increases by about 43%, while BSA and estimated FFM increase by about 21% and 24%. Under the stipulated exponents, weight therefore produces the largest predicted increases in both parameters. That arithmetic alone does not show that weight is the most appropriate descriptor.

Worked Example: Same Weight, Different Estimated FFM

Suppose two hypothetical adults, one male and one female, each weigh 90 kg and are 175 cm tall. Their weight and BSA are the same: \(BSA=\sqrt{175(90)/3600}=\sqrt{4.375}=2.092\ \mathrm{m^2}\). Their BMI is \(90/1.75^2=29.39\ \mathrm{kg/m^2}\).

Using the stated adult equations, the male estimate is \(FFM=9270(90)/[6680+216(29.39)]=834300/13028=64.04\ \mathrm{kg}\). The female estimate is \(FFM=9270(90)/[8780+244(29.39)]=834300/15951=52.30\ \mathrm{kg}\). The estimated-FFM ratio is \(64.04/52.30=1.224\), whereas the weight and BSA ratios are both 1.

If, solely to illustrate the scaling consequence, a clearance model were anchored at 5.00 L/h for the female and used an FFM exponent of 0.75, the male prediction would be \(5.00(1.224)^{0.75}=5.82\ \mathrm{L/h}\). A weight-based model with the same anchor and exponent would predict \(5.00(1)^{0.75}=5.00\ \mathrm{L/h}\). This contrast comes from the specified FFM equations; it is not evidence that the estimates capture all biologically relevant differences between these hypothetical individuals.

Worked Example: Equal Weight Does Not Mean Equal Scaling

Consider two hypothetical men who each weigh 90 kg. The shorter man is 160 cm tall and the taller man is 190 cm tall. Weight scaling gives a descriptor ratio of \(90/90=1\), so it predicts no size-related difference between them under a model anchored at either person.

Their BSA values are \(\sqrt{160(90)/3600}=\sqrt{4}=2.000\ \mathrm{m^2}\) and \(\sqrt{190(90)/3600}=\sqrt{4.75}=2.179\ \mathrm{m^2}\). The taller-to-shorter BSA ratio is \(2.179/2.000=1.090\). Their BMIs are \(90/1.60^2=35.16\) and \(90/1.90^2=24.93\ \mathrm{kg/m^2}\). Applying the male FFM equation gives \(FFM_{\mathrm{short}}=834300/[6680+216(35.16)]=58.45\ \mathrm{kg}\) and \(FFM_{\mathrm{tall}}=834300/[6680+216(24.93)]=69.15\ \mathrm{kg}\), an FFM ratio of \(69.15/58.45=1.183\).

With an assumed clearance exponent of 0.75 and an anchor clearance of 5.00 L/h for the shorter man, BSA scaling predicts \(5.00(1.090)^{0.75}=5.33\ \mathrm{L/h}\), while FFM scaling predicts \(5.00(1.183)^{0.75}=5.67\ \mathrm{L/h}\). Weight scaling predicts 5.00 L/h. This example shows how a descriptor can encode height or estimated composition even when total weight is unchanged.

Why the Ratio Formula Works

The calculation used above follows directly from the power-law model. If \(P_A=kS_A^\theta\) and \(P_B=kS_B^\theta\), divide the second equation by the first. The common constant cancels:

\frac{P_B}{P_A} = \frac{kS_B^\theta}{kS_A^\theta} = \left(\frac{S_B}{S_A}\right)^\theta.

This also explains why the descriptor's units do not create a problem for the ratio: \(S_B/S_A\) is dimensionless, provided both values use the same units. The absolute model constant \(k\) would depend on the units and parameterization, but it cancels in this comparison. The formula requires the same \(k\) and \(\theta\) for the two individuals within the candidate model. It does not imply that different descriptors should share the same fitted constant in an unrestricted model.

Choosing and Checking a Descriptor

A defensible choice begins with the application and data, not with a universal ranking of descriptors. Weight may be preferred when it is reliably recorded and the population resembles the data supporting the model. BSA may be useful when both height and weight are available and the geometric combination is scientifically relevant. FFM may be worth evaluating when differences in body composition are important, but it depends on an estimation equation and can carry its own uncertainty. A descriptor that is unavailable or unreliable in the intended population may be a poor operational choice even if its biological rationale is attractive.

If alternative descriptors are strongly correlated in the study population, they may yield similar predictions over the observed range. That makes the choice difficult to distinguish from those data, not necessarily irrelevant at more extreme sizes. Conversely, as the examples show, the descriptors can diverge for particular body types. A practical comparison should examine predictions across the actual covariate range and at clinically relevant extremes, rather than only compare a single fitted coefficient.

Useful checks include inspecting observed-versus-predicted concentrations or parameter-related diagnostics across descriptor values, looking for systematic residual patterns by weight, height, or estimated FFM, and assessing predictive performance in data not used for fitting when available. A descriptor can appear adequate in a narrow size range yet extrapolate poorly. Since BSA and FFM are calculated from measured inputs, errors in height, weight, or the FFM equation also affect the resulting descriptor; treating it as exact may understate uncertainty.

State the candidate descriptor.
Specify whether size is represented by weight, BSA, FFM, or another quantity, including units and any equation used to calculate it.

Hold other choices clear.
When isolating the descriptor's impact, state the exponent and common anchor parameter. In an actual model comparison, do not assume that equal exponents or fit quality follow automatically.

Check fit and predictive behavior.
Examine residual patterns and predictions over the observed and intended-use ranges. Check whether the preferred descriptor is supported by data, not just by plausibility.

Common Pitfalls and Limitations

Assuming that kilograms, square metres, and FFM kilograms are interchangeable. They are different descriptors. Their ratios across individuals can differ, so their allometric predictions can differ.
Comparing raw descriptor values without units or a defined equation. BSA depends on the chosen formula, and FFM depends on its estimation method. State these choices before interpreting results.
Calling one descriptor universally mechanistic. Weight, BSA, and FFM are proxies for complex physiology. A plausible proxy does not guarantee better fit or prediction for a particular drug.
Changing descriptor and exponent together, then attributing the result to the descriptor alone. The worked comparisons held the exponent fixed to isolate one source of change. A fitted analysis should report all model choices and compare their performance.
Extrapolating beyond the data without checking. Similar predictions among descriptors over a limited range do not ensure similar predictions for very small, very large, or compositionally unusual individuals.

The main limitation is that descriptor choice is partly an empirical and application-specific decision. The equations in the examples demonstrate the consequences of alternatives; they do not determine the best choice for every parameter. A useful model must be interpretable, calculable for the intended population, and supported by adequate data over the range where predictions will be used.

Check Your Understanding

Answer each question, then compare your reasoning with the solutions.

For a common anchor value \(P_A\), what is the predicted ratio \(P_B/P_A\) under a power-law model with descriptor \(S\) and exponent \(\theta\)?
Two people have the same weight but different heights. Which of weight, BSA, and FFM can distinguish them in general, and why?
Under a clearance exponent of 0.75, a descriptor ratio is 1.20. What is the predicted clearance ratio, approximately?
Name two checks that can reveal whether a descriptor is inadequate for the model's intended use.

Solutions. (1) \(P_B/P_A=(S_B/S_A)^\theta\), obtained by dividing the two power-law equations. (2) Weight cannot distinguish them because its values are equal. BSA can differ because it uses height and weight, and estimated FFM can differ because the equation uses BMI and weight. (3) \(1.20^{0.75}=\exp[0.75\ln(1.20)]\approx1.147\), so the predicted clearance is about 14.7% higher for the person with the larger descriptor, under the stated model. (4) Examples include checking residual patterns across size measures, evaluating predictions over the intended-use range, and assessing predictive performance in data not used for fitting.