Selecting Body-Size Descriptors for Obese Adults

Why the Descriptor Matters

A PBPK model for an obese adult needs physiological inputs such as organ volumes and blood flows. A model may use body size to scale those inputs from a reference subject, but “body size” is not a single quantity. Total body weight (TBW), ideal body weight (IBW), lean body mass (LBM), and adjusted body weight (AdjBW) are different descriptors, and they need not imply the same prediction for a given organ or flow.

The central modeling decision is not simply which descriptor gives a convenient number. It is which descriptor, if any, is supported for the particular physiological quantity and population. A descriptor used to calculate a covariate is not itself proof that the covariate changes in direct proportion to that descriptor. The equations below make alternative assumptions explicit so their consequences can be compared.

Defining the Common Descriptors

One commonly used adult IBW convention is the Devine formula. With height \(H\) in inches, the formula is \(50+2.3(H-60)\) kg for men and \(45.5+2.3(H-60)\) kg for women, when \(H\) is at least 60 inches. These are empirical reference formulas; they do not define a biological boundary between lean and adipose mass. Other IBW conventions exist, so a model should record which formula it uses.

A commonly used LBM estimate for adults is the Janmahasatian equation. It uses TBW \(W\) in kilograms and body mass index, \(BMI=W/H^2\), where height \(H\) is in metres. The sex-specific equations are:

These are prediction equations, not direct body-composition measurements. Their use should be documented, particularly when applying them outside the populations in which they are suitable. A common adjusted-weight definition is \(AdjBW=IBW+0.4(TBW-IBW)\). The factor 0.4 is a convention, not a universal physiological constant. If a different fraction is selected, it should be stated and justified.

Using a Descriptor in a Scaling Relationship

A body-size descriptor enters a PBPK model through a specified relationship between a reference physiological quantity and the corresponding quantity in a subject. For a quantity \(Y\), a reference value \(Y_{\mathrm{ref}}\), and a descriptor \(D\), a simple power relationship is:

Here \(D_{\mathrm{ref}}\) is the descriptor value for the reference subject, and \(b\) is the scaling exponent. The ratio is dimensionless when \(D\) and \(D_{\mathrm{ref}}\) use the same units, so \(Y\) retains the units of \(Y_{\mathrm{ref}}\). With \(b=1\), the model assumes direct proportionality to the selected descriptor. With another exponent, it assumes a different relationship. Neither the descriptor nor the exponent should be chosen merely because the equation is easy to calculate.

The logic of the relationship follows by assuming that \(Y/D^b\) is constant between the subject and reference. Equating those ratios gives \(Y/D^b=Y_{\mathrm{ref}}/D_{\mathrm{ref}}^b\); multiplying both sides by \(D^b\) yields the displayed scaling equation. This derivation also makes the assumption visible: the same proportionality constant and exponent are being applied to both subjects.

Comparing Predictions Without Treating Them as Evidence

A useful first comparison holds the reference value and exponent fixed while changing only the descriptor. This isolates the consequence of the descriptor choice. For example, suppose a hypothetical organ volume is \(1.50\) L at \(D_{\mathrm{ref}}=70\) kg and the model uses \(b=1\). For the hypothetical man in the earlier calculation, the four linear predictions are:

This spread is a sensitivity result: the predictions depend strongly on the selected descriptor under the imposed linear relationship. It does not establish which prediction is physiologically correct. In particular, total body mass includes mass that need not produce a proportional increase in every organ volume or flow. Conversely, using IBW or LBM does not guarantee that an organ or flow remains unchanged as obesity increases.

The same comparison can be made for other physiological inputs, but each needs its own justification. Organ volume and blood flow are distinct quantities and need not share a descriptor or exponent. A single body-size formula applied to all organs and flows may be computationally consistent yet physiologically unsupported.

Assumptions, Checks, and Alternatives

Before applying a scaling relationship, state the assumptions that make its interpretation possible. The reference and subject must be described using the same definition and units of the descriptor. The reference physiological value must correspond to the intended reference population and model quantity. The relationship must be appropriate over the range of sizes being considered, and any chosen exponent must be supported or explicitly treated as an assumption. If the model uses a size descriptor as a proxy, it assumes that this proxy captures the relevant variation for that specific input.

Diagnostics should examine both calculations and model behavior. First verify units, height conversions, sex-specific formulas, and arithmetic. Then plot or tabulate each input against TBW and the alternative descriptors across the virtual cohort. Look for abrupt changes, implausible values, and systematic trends in the model outputs that arise only from a descriptor choice. Compare the predicted inputs with available physiological constraints or observations where they exist. If only one descriptor yields plausible predictions, that is a reason to investigate the model assumptions, not automatically proof that the descriptor is correct.

The main alternatives are not simply four competing weight formulas. A model may use an organ-specific empirical relationship, a directly measured physiological value, or a reference value without body-size scaling when that is appropriate. It may also compare multiple prespecified descriptor-based models as a sensitivity analysis. These approaches answer different questions: a direct measurement supplies an input, an empirical relationship estimates a quantity from data, and a descriptor-based rule extrapolates from a reference under an assumed form. The choice should follow the available evidence and the modeling purpose.

In constructing an obese-adult cohort, body-size descriptors should be calculated consistently for each virtual subject. Earlier material on sampling continuous physiological variability and preserving dependence among covariates is relevant here: height, weight, and derived descriptors are not independent quantities. Calculating multiple descriptors does not justify sampling each as though it were an unrelated covariate.

Common Pitfalls and Limits

• Treating AdjBW as a biological measurement. It is a constructed descriptor, and its correction fraction is conventional. State the formula rather than implying that the result is measured lean tissue.
• Assuming “ideal” means physiologically correct. IBW is a height-based estimate. It does not identify the appropriate scaling variable for an organ or flow.
• Applying one descriptor to every parameter. A relationship that is reasonable for one physiological quantity is not automatically appropriate for another.
• Changing descriptor and exponent together without explanation. The resulting change in prediction is then difficult to interpret. Vary one choice at a time for diagnosis.
• Scaling beyond the evidence. A formula that behaves smoothly over its calculation range can still be an unsupported extrapolation to severe obesity or a different population.
• Double-counting body size. If body size has already entered a physiological input through a descriptor-based scaling rule, adding a second adjustment for the same size effect may inflate the change unless the model explicitly represents distinct mechanisms.

These descriptors are useful for organizing candidate models and quantifying sensitivity, but they do not resolve the underlying physiological question. The appropriate choice can depend on the organ, flow, population, available measurements, and intended extrapolation. Later modeling steps can represent obesity-specific changes in particular tissues and flows; those changes should not be presumed from a single whole-body weight scalar.