Weight Effects on Absorption and Distribution

Why Assess Absorption and Distribution Separately?

A relationship between body weight and drug concentrations does not, by itself, show which pharmacokinetic parameter varies with weight. Concentrations after an oral dose reflect both absorption into the systemic circulation and subsequent distribution and elimination. Intravenous data bypass absorption, but still reflect distribution and elimination. The analysis therefore needs to ask a parameter-specific question: is weight associated with the absorption rate, the fraction reaching systemic circulation, one or more distribution volumes, or some combination?

The distinction matters clinically and statistically. A weight effect on absorption rate can change the timing and peak of concentrations without changing total exposure, under linear pharmacokinetics and complete observation of the concentration-time curve. A change in bioavailability can change total exposure. A change in distribution volume affects the concentration associated with a given amount of drug in the body and can also affect the concentration-time profile. These effects should not be treated as interchangeable explanations for the same pattern.

For a positive parameter \(P_i\) in individual \(i\), a common way to describe a weight relationship is a power function centered at a reference weight:

Here \(WT_i\) is the individual's weight, \(WT_{\mathrm{ref}}\) is a chosen reference weight, \(P_{\mathrm{ref}}\) is the typical parameter value at that reference, and \(\theta_P\) is the parameter-specific weight exponent. An exponent of zero represents no weight relationship in this model. A positive exponent means the parameter increases with weight; a negative exponent means it decreases. The exponent for \(k_a\) need not equal the exponent for a distribution volume. The earlier material on power-function models and allometric scaling provides the general covariate framework; the central issue here is choosing which parameter, if any, should receive a weight relationship.

What the Route of Administration Can Tell You

After an intravenous dose, absorption parameters do not govern the concentration-time profile. In a one-compartment model with an intravenous bolus dose \(D\), the initial concentration is \(C(0)=D/V\), where \(V\) is the distribution volume. Under linear elimination, the concentration subsequently declines according to the elimination rate. Thus, sufficiently informative intravenous data can support assessment of distribution and elimination parameters without confounding them with oral absorption.

For an oral dose, the observed profile depends on the input process as well as disposition. In a simple one-compartment model with first-order absorption and elimination, \(k_a\) influences the rate of rise and the timing of the peak. If pharmacokinetics are linear and the full concentration-time curve is observed, the area under the curve (AUC) is \(F D/CL\), where \(CL\) is clearance. This means AUC alone does not separate bioavailability from clearance. Likewise, oral-only data commonly identify apparent parameters such as \(CL/F\) and \(V/F\), rather than \(CL\), \(V\), and \(F\) separately.

The model assumptions must be considered before interpreting a fitted relationship. The power function assumes a smooth, monotonic relationship over the observed weight range. Linear AUC arguments assume dose-proportional pharmacokinetics and appropriate AUC measurement. A first-order absorption model assumes the input rate is proportional to the amount remaining to be absorbed. In practice, formulation, food, dosing conditions, adherence, and sampling times can also affect oral profiles. If any of these vary systematically with weight, an apparent weight effect may partly reflect those factors.

Assess Absorption Parameters on Their Own Evidence

A weight effect on \(k_a\) is most directly supported by oral concentration-time data that adequately capture the absorption phase. Sparse sampling near and after the expected peak may leave \(k_a\) weakly estimated. If absorption is slower than elimination, the terminal part of an oral profile can reflect absorption rather than elimination, a situation often called flip-flop kinetics. In such settings, a fitted weight relationship for \(k_a\) may compensate for an incorrect structural model or insufficient information about the terminal phase.

Bioavailability \(F\) is a different question from absorption rate. It describes the fraction of the dose reaching systemic circulation, not how quickly that fraction arrives. In an oral-only study, a weight relationship in \(F\) may be difficult to distinguish from a relationship in clearance because AUC depends on \(F/CL\). Comparing dose-normalized oral and intravenous AUCs can estimate absolute bioavailability when the routes are studied under comparable conditions and linear pharmacokinetics is reasonable:

This comparison does not establish a weight effect unless the study provides suitable observations across weight and accounts for relevant differences in conditions. Nor does it identify a weight effect on \(k_a\): AUC is an exposure measure, not a direct measure of absorption timing.

Separate Bioavailability From Disposition When Possible

The dose-normalized AUC comparison gives a direct way to illustrate why oral and intravenous data can lead to different conclusions. For each route, the AUC reflects both the amount entering systemic circulation and systemic clearance. With an intravenous dose, \(F=1\) by definition for the administered dose; with an oral dose, the ratio of dose-normalized AUCs estimates the fraction reaching systemic circulation under the assumptions above.

Model Choice, Diagnostics, and Limitations

A defensible analysis begins with a disposition and absorption model that is adequate before adding weight effects. Then assess candidate effects separately: for example, weight on \(k_a\), on \(F\), and on each volume relevant to the structural model. A distribution model may contain more than one volume; evidence for a weight effect on one volume does not automatically justify the same effect on another. Likewise, a weight effect on \(F\) does not imply a weight effect on absorption rate.

Inspect both parameter estimates and model predictions. Useful checks include whether individual parameter estimates show a systematic trend with weight, whether the fitted exponent is precise enough to support its interpretation, and whether predictions reproduce the observed concentration-time profiles at low and high weights. For oral data, pay particular attention to the absorption phase and peak; for intravenous data, check early distribution and the later elimination profile. A relationship that improves one part of the profile while worsening another may indicate a misspecified model rather than a useful covariate effect.

Common warning signs include a large change in the weight exponent under modest changes to the model, high uncertainty in the exponent, implausible parameter predictions outside the observed weight range, or a trend in residuals that remains after adding weight. In oral-only data, strong dependence between \(F\) and \(CL\), or between \(F\) and \(V\), may make separate weight effects weakly identified. Weight may also be associated with age, body composition, disease severity, dose, or administration conditions. A fitted association should not be interpreted as a causal effect of weight without a design and model that support that interpretation.

The main alternatives answer different questions. A model with no weight effect is the simplest benchmark and may be preferable when the data do not support a stable relationship. A shared weight exponent across several parameters is more parsimonious, but it imposes a biological and statistical constraint that may be inappropriate for absorption and distribution. Separate parameter-specific exponents allow distinct patterns but require more information and can be difficult to estimate. A categorical weight grouping or a more flexible relationship can reveal departures from a simple power function, but may use data inefficiently or overfit when observations are limited. Model choice should reflect the scientific question, observed weight range, parameter identifiability, and predictive performance—not a preference for adding covariates.