Why Scaling Errors Can Produce Plausible but Wrong Predictions
A pediatric clearance prediction can be wrong even when every component of its equation looks familiar. The common source is not arithmetic but accounting: body growth may be included twice, adult enzyme activity may be treated as established in a child without evidence, or an age-dependent function may be evaluated using an age definition different from the one used to estimate it. Such errors can yield numbers that look biologically reasonable while systematically understating or overstating clearance.
The central audit question is: what biological quantity does each factor scale, and has that quantity already been represented elsewhere? The answer depends on model structure. A whole-body empirical allometric relationship and an organ-resolved PBPK model can both be defensible, but their scaling factors should not be combined mechanically.
Keep Whole-Body and Organ-Level Scaling Separate
A common whole-body clearance model is
where \(W\) is body weight and \(MF(\mathrm{age})\) is a maturation factor for the clearance pathway being modeled. The exponent 0.75 is a model assumption commonly used in some allometric approaches; it is not a universal biological constant. The maturation factor must be defined relative to the reference activity implicit in \(CL_{\mathrm{adult}}\). If it is a fraction of adult activity, it is dimensionless and equals 1 at the adult reference age by the chosen normalization.
In an organ-resolved model, the accounting may instead be explicit. For example, intrinsic clearance can depend on organ mass, enzyme abundance per unit organ mass, and activity per unit enzyme. A general proportional representation is
where \(M_{\mathrm{organ}}\) is organ mass, \(A_{\mathrm{enzyme/organ}}\) is enzyme abundance per unit organ mass, and \(k_{\mathrm{activity}}\) represents catalytic activity per unit enzyme under the relevant conditions. These quantities are not interchangeable. Body weight may inform organ mass, while ontogeny data may inform enzyme abundance; a change in catalytic activity is a separate claim requiring its own evidence.
This factorization gives a practical test for double counting. If the model already scales liver mass with age and scales enzyme abundance per gram of liver with a maturation function, multiplying the resulting organ-level intrinsic clearance by a second, whole-body clearance allometric factor may count growth again. Conversely, if the prediction begins with adult whole-body clearance and uses one empirical allometric factor to represent size scaling, adding a separate liver-mass ratio may duplicate part of that factor. The correct choice depends on the defined model, not on a rule that one particular factor must always be present.
Worked Example: Detecting Double-Counted Growth
Suppose a hypothetical adult has \(CL_{\mathrm{adult}}=100\) L/h at \(W_{\mathrm{adult}}=70\) kg. A child weighs 10 kg. For this example, assume that the selected whole-body model uses an exponent of 0.75 and that the pathway maturation factor is 0.50. These are stipulated modeling assumptions, not clinical estimates.
The weight ratio is \(10/70=0.1429\). The size factor is
Using the stated whole-body model gives
Now suppose someone additionally multiplies by a hypothetical liver-mass ratio of 0.25, despite using an allometric factor already intended to scale whole-body clearance with size. That calculation gives \(100\times0.2324\times0.50\times0.25=2.905\) L/h. The extra factor reduces the prediction fourfold. It is not justified by the information given: in this model, it risks counting growth twice.
An organ-resolved alternative could use the liver-mass ratio and enzyme abundance per gram of liver explicitly, but then it would need a coherent organ-level clearance calculation rather than automatically retaining the whole-body clearance allometry. The example diagnoses a structural problem; it does not establish that 11.62 L/h is the clinically correct value.
Do Not Assume Adult Activity Without a Defined Reference
A maturation factor is meaningful only when its biological target and reference are explicit. If \(MF\) represents enzyme abundance relative to adult abundance, it should not silently be interpreted as a correction for all developmental differences in hepatic clearance. Binding, blood flow, competing pathways, and organ size may be represented elsewhere in the PBPK model. The earlier material on integrating age-dependent hepatic intrinsic clearance distinguishes pathway-specific capacity from whole-organ clearance; that distinction matters when auditing a scaling equation.
Assuming \(MF=1\) at every pediatric age amounts to assuming adult-level activity for the modeled quantity. That may be a useful hypothesis for a pathway known to be mature in the relevant age range, but it is not a neutral default. Likewise, setting a low maturation factor based only on age, without considering the enzyme or transporter and the evidence behind its ontogeny function, can impose an unsupported developmental reduction.
Worked Example: Quantifying an Adult-Activity Assumption
Consider a hypothetical adult with clearance 60 L/h at 70 kg and a hypothetical child weighing 15 kg. Use the same stipulated allometric exponent of 0.75. Suppose the evidence-supported maturation factor for the pathway at the child’s age is assumed, for this calculation, to be 0.30.
The size factor is
The prediction that includes the stipulated maturation factor is
If adult activity is assumed by setting \(MF=1\), the prediction instead becomes \(60\times0.3150=18.90\) L/h, which is 3.33 times the value using \(MF=0.30\). The difference follows directly from the assumption, not from a change in body weight. Neither result is supported merely by the arithmetic: the analyst must justify whether the maturation factor applies to this pathway and age, and whether the allometric model and its reference clearance are suitable.
Age Definitions Must Match the Maturation Function
An age-dependent function is calibrated for a particular age variable and unit. Depending on the purpose and available evidence, that variable may be chronological postnatal age, gestational age, postmenstrual age, or another explicitly defined scale. Postmenstrual age is gestational age at birth plus elapsed postnatal age, expressed in a consistent unit. These ages are not numerically interchangeable. A maturation function estimated against postmenstrual age cannot be evaluated using postnatal age alone without changing the model.
For illustration, a commonly used saturating form is
where \(a\) is the specified age variable, \(TM_{50}\) is the value of that same age variable at which the function reaches 0.5, and \(H\) is a shape parameter. The units for \(a\) and \(TM_{50}\) must agree. This function is not automatically appropriate for every pathway or population; its structure and parameters require justification.
Worked Example: Correcting an Age-Scale Mismatch
Suppose a hypothetical maturation function is defined using postmenstrual age in weeks, with \(TM_{50}=40\) weeks and \(H=2\). Consider an infant born at 32 weeks’ gestational age and evaluated 8 weeks after birth. The postmenstrual age is \(32+8=40\) weeks, so
If the analyst mistakenly inserts postnatal age, the result is
The second calculation is arithmetically correct but uses the wrong age variable for this stipulated function. To show the impact on a hypothetical clearance prediction, take \(CL_{\mathrm{adult}}=100\) L/h at 70 kg, child weight 5 kg, and the same 0.75 allometric exponent. The size factor is \((5/70)^{0.75}=0.1382\). Using the correctly defined maturation factor gives \(100\times0.1382\times0.5000=6.910\) L/h. Using the mismatched factor gives \(100\times0.1382\times0.03846=0.5315\) L/h. The large discrepancy is caused by inconsistent age definition, not by a subtle numerical effect.
Extrapolation Is Not Validation
A maturation equation can always return a number outside the age range that informed its parameters. That numerical output does not show that the function is valid there. Extrapolation is especially consequential near birth, in prematurity, or at ages where the function changes steeply. Check the range and age definition of the supporting data, the biological plausibility of the curve at the boundary, and how uncertainty in parameters affects the prediction. The earlier material on propagating ontogeny uncertainty is relevant when such parameter uncertainty is material to exposure predictions.
Do not repair a lack of evidence by silently changing the age scale, forcing a boundary value, or treating an extrapolated point estimate as established. Instead, state the unsupported range, show the assumption used, and consider scenario or sensitivity analyses. A model may still be useful for exploring possibilities, but its prediction should be labeled as uncertain and not presented as observed performance.
Worked Example: Showing Sensitivity in an Unsupported Age Range
Suppose a hypothetical maturation function has the same form with \(H=2\), but supporting data cover postmenstrual ages from 28 to 52 weeks. A prediction at 8 weeks is outside that range. To illustrate parameter sensitivity, compare two hypothetical values, \(TM_{50}=30\) weeks and \(TM_{50}=50\) weeks, while keeping the function form and age variable fixed.
At 8 weeks, the first parameter set gives
The second gives
The first value is about \(0.06639/0.02496=2.660\) times the second. This calculation does not establish which \(TM_{50}\) is correct; it shows that an out-of-range prediction can be sensitive to a parameter that the available data may not determine well. Reporting only one extrapolated value would conceal that dependence.
A Practical Audit and Its Alternatives
Before accepting a pediatric scaling result, trace every factor from the adult reference to the predicted child quantity. Check units, reference age and weight, the meaning of the maturation factor, and whether the age variable matches the function’s definition. Then examine whether the predicted clearance trend is compatible with the intended model and available observations. A mismatch is a prompt to investigate structure and evidence, not to tune factors until the curve looks plausible. The earlier material on diagnosing confounding between age, size, and maturation and on checking pediatric predictions against observed clearance trends provides useful context for these checks.
State whether each equation predicts whole-body clearance, intrinsic clearance, enzyme abundance, organ mass, or another parameter.
Explain which change is represented by body-weight allometry, organ scaling, maturation, or an activity term.
Verify the adult reference, age variable, units, and calibration range for every factor.
Check sensitivity, compare with suitable observed trends, and label extrapolations and unsupported assumptions.
The main alternatives differ in what they model directly. A simple allometric approach can be transparent and practical when its reference and domain are appropriate. An organ-resolved PBPK approach can represent growth and pathway ontogeny separately, but requires corresponding physiological and biochemical inputs. A maturation function adds an explicit age-dependent component, but its parameters and age scale must be supported. These approaches are not interchangeable labels for the same calculation; a model should use the structure that matches the available evidence and intended prediction.
| Approach | What it represents | Audit question |
|---|---|---|
| Whole-body allometry | Size dependence of a specified whole-body parameter | Is the exponent and reference relationship appropriate, and is organ growth being added again? |
| Organ-resolved scaling | Physiological organ size and organ-specific capacity | Are organ mass, abundance per organ mass, and activity kept distinct? |
| Age-dependent maturation | Change in a defined biological capacity with age | Does the age definition match the data, and is the prediction within the supported range? |
Common Mistakes and Limitations
- Multiplying every available factor. More factors do not automatically make a model more mechanistic. Include a factor only when its role is defined and not already captured elsewhere.
- Calling a maturation factor “clearance.” A factor for enzyme abundance is not necessarily a factor for whole-organ clearance, which can also depend on organ physiology and other pathways.
- Assuming adult activity by default. \(MF=1\) is an assumption about the modeled capacity, not a consequence of using adult clearance as the reference.
- Changing age units or definitions unnoticed. Weeks of postnatal age and weeks of postmenstrual age are not interchangeable, even though both are reported in weeks.
- Interpreting an extrapolated output as validated. A smooth curve can produce a precise number far outside the data range; numerical precision is not evidence of predictive accuracy.
No audit can substitute for appropriate pathway-specific data, and a structurally consistent model can still be uncertain because of parameter estimates, sparse pediatric observations, or unmodeled biology. The aim is narrower but essential: make the assumptions visible, prevent redundant scaling, and ensure that the calculated prediction means what the model claims it means.
Check Your Understanding
Answer each question, then compare your reasoning with the solutions.
- A whole-body clearance allometry is used, and a liver-mass ratio is multiplied in afterward. What must be established before concluding that both factors belong in the model?
- A maturation function uses postmenstrual age in weeks. An infant was born at 30 weeks’ gestational age and is now 6 weeks postnatal. What age should be entered?
- For \(MF(a)=a^2/(40^2+a^2)\), calculate \(MF(20)\).
- What does a maturation factor of 1 mean if the function is normalized to adult activity? Does it prove adult whole-body clearance?
- A parameterized maturation function is used well below the ages represented in its supporting data. Name two appropriate ways to report or investigate the result.
Solutions. (1) Establish that the allometric factor does not already represent the growth captured by the liver-mass ratio, and that both factors apply to the defined target quantity. (2) Enter \(30+6=36\) weeks postmenstrual age. (3) \(MF(20)=20^2/(40^2+20^2)=400/2000=0.2000\). (4) It means the modeled activity equals the function’s adult reference level; it does not by itself establish adult whole-body clearance, which may depend on other physiological and clearance components. (5) State that the prediction is extrapolated and report sensitivity or scenario results across plausible parameter values; do not describe the output as validated solely because the equation returns a number.