How Allometry Changes Half-Life Across Body Sizes

Why Half-Life Has Its Own Size Dependence

Allometric models describe how pharmacokinetic parameters change with body size, but half-life is not itself an independent parameter in the simplest elimination model. It is determined jointly by clearance and volume of distribution. If those two quantities scale differently with body size, their ratio changes too, and so does the time required for concentration to fall by half.

The key result is simple: the body-size exponent for half-life is the volume exponent minus the clearance exponent. This difference determines whether the model predicts a longer, shorter, or unchanged half-life as body size increases. It does not, by itself, establish that a particular exponent is appropriate for a drug or population; the exponents and model assumptions still require justification.

Deriving the Half-Life Exponent

Let \(W_{\mathrm{ref}}\) be a reference weight, and let \(CL_{\mathrm{ref}}\) and \(V_{\mathrm{ref}}\) be clearance and volume at that weight. Under power-law allometry, the typical parameter values at weight \(W_i\) are

For the one-compartment first-order model, the concentration elimination rate constant is \(k_i=CL_i/V_i\). The half-life is the time \(t\) for which the exponential concentration decline satisfies \(e^{-k_i t}=1/2\). Taking logarithms gives \(k_i t=\ln(2)\), so \(t_{1/2,i}=\ln(2)/k_i=\ln(2)V_i/CL_i\). Substituting the allometric expressions yields

where \(t_{1/2,\mathrm{ref}}=\ln(2)V_{\mathrm{ref}}/CL_{\mathrm{ref}}\). Thus the half-life exponent is a difference, not either parameter’s exponent taken alone. If the exponents are equal, their size effects cancel in the ratio and the model predicts a size-invariant half-life. If \(\beta_V>\beta_{CL}\), half-life increases with weight; if \(\beta_V<\beta_{CL}\), it decreases.

In an NLME model, an individual may also differ from the typical values through random effects. For example, if \(CL_i\) and \(V_i\) include multiplicative factors \(e^{\eta_{CL,i}}\) and \(e^{\eta_{V,i}}\), then the individual half-life includes the factor \(e^{\eta_{V,i}-\eta_{CL,i}}\), as well as the weight term. The exponent difference describes the typical size trend when the random effects are held fixed; individuals of the same weight can still have different predicted half-lives.

Interpreting the Exponent Difference

The exponent difference is a relative scaling statement. It describes how the ratio \(V/CL\) changes when weight changes, not how either parameter behaves in isolation. For a weight ratio \(r=W_i/W_{\mathrm{ref}}\), the predicted half-life ratio is \(r^{\beta_V-\beta_{CL}}\). A positive difference means a larger body size is associated with a longer model-predicted half-life, all else held constant. It does not imply that body size causes every individual’s half-life to change by that amount.

Estimating or Checking a Half-Life Scaling Trend

If half-lives are calculated from a fitted clearance and volume model, their typical size trend follows from the difference in fitted exponents. Conversely, if exact model-predicted half-lives at two weights are available and one parameter exponent is known, the other can be recovered algebraically. With \(r=W_2/W_1\), the relation is

This inversion is a mathematical identity for the assumed power law. In real data, observed half-lives contain estimation error and may not follow an exact power law, so the formula should not be treated as a substitute for fitting and evaluating an appropriate model.

Diagnostics, Alternatives, and Limitations

A useful check is to calculate half-life from the model’s clearance and volume at several weights and compare those calculations with the exponent-difference formula. Disagreement usually indicates an arithmetic, parameterization, or implementation error. In an NLME analysis, inspect whether the fitted clearance and volume relationships are plausible over the observed weight range, and whether individual half-life predictions show systematic patterns by weight. A trend in residuals or unexplained weight-related structure can signal that the assumed exponents or model form do not represent the data adequately. These checks do not prove that the model is correct; they help reveal incompatibilities and influential extrapolations.

The half-life expression also has an important boundary. In a multi-compartment model, \(CL/V\) generally does not determine every concentration decline rate, and the observed terminal half-life depends on distribution between compartments as well as elimination. The simple formula should therefore not be applied to a terminal half-life in a multi-compartment model without a derivation appropriate to that model. Intercompartmental clearance and distribution can affect the terminal phase even when systemic clearance is unchanged.

An alternative to deriving half-life as a ratio is to calculate individual \(CL_i\) and \(V_i\) from their allometric models and then compute \(\ln(2)V_i/CL_i\). Under the assumptions above, this gives the same result and is a useful numerical verification. Modeling a half-life trend directly can be convenient descriptively, but it does not automatically preserve the relationship between clearance and volume or their uncertainty. For mechanistic interpretation, deriving the trend from the two parameter models makes the assumptions transparent.

The result is limited by the model and the weight range. Power-law scaling may be a reasonable approximation over a particular range but should not be assumed to extrapolate indefinitely. Also, a population-level exponent describes a typical trend: random effects, other covariates, and parameter uncertainty can produce substantial variation around it. Finally, half-life is not interchangeable with exposure. Clearance is central to exposure after a dose, while half-life depends on both clearance and volume.

Common Pitfall

A frequent error is to state that half-life scales with the clearance exponent, or to use the clearance exponent alone because clearance is associated with elimination. The half-life formula shows why this is wrong: \(t_{1/2}=\ln(2)V/CL\), so both exponents enter, with the clearance exponent subtracted. For example, if \(\beta_V=1.00\) and \(\beta_{CL}=0.75\), half-life scales with exponent \(0.25\), not \(0.75\) or \(1.00\). Another error is to conclude that a positive exponent guarantees a longer observed terminal half-life in a multi-compartment system; the simple derivation applies to the specified one-compartment model, not automatically to a terminal phase.