Why Choose a Reference Age?
A maturation curve can be expressed as a fraction of a mature plateau, or as a multiplier anchored to a particular age. These choices answer different practical questions. A plateau-based parameterization asks how far a patient’s maturation has progressed toward the mature value. A reference-age parameterization asks how the value at another age compares with a specified age, such as one year.
The reference age is a coordinate choice: it can make a model parameter correspond to a meaningful clinical quantity, such as typical clearance at a particular age. Provided the parameters are converted consistently, changing the reference does not change the predicted curve. It does change what the parameters mean, so the reference age and its units must be stated explicitly.
Plateau-Based and Reference-Age Parameterizations
Let \(a\) denote age on a specified age scale, \(a_{50}\) the half-maturation age, and \(h\) the Hill coefficient. The standard sigmoidal Hill maturation fraction is
This curve rises from 0 at age 0 toward a mature plateau of 1. At \(a=a_{50}\), \(F(a)=0.5\). If a quantity has a mature plateau \(M_{\infty}\), its age-dependent value can be written \(M(a)=M_{\infty}F(a)\). Here \(M_{\infty}\) is the value approached as age increases; it is not the value at every age in the observed data.
For a normalized Hill fraction, define \(F_{\mathrm{ref}}=F(a_{\mathrm{ref}})\). The relative multiplier is
It equals 1 at \(a_{\mathrm{ref}}\), but its mature plateau is \(1/F_{\mathrm{ref}}\), which is generally greater than 1. Thus a value normalized to a young reference age should not be mistaken for a fraction bounded above by 1. The plateau-based fraction \(F(a)\) is bounded by 1; the reference-normalized ratio \(R(a;a_{\mathrm{ref}})\) is a comparison with the reference value.
Converting a Reference Value to Hill Parameters
A model can use \(M_{\mathrm{ref}}\) and \(M_{\infty}\) directly instead of using \(a_{50}\) to describe the curve’s location. This is useful when the quantity at a specific age is more interpretable than the age at which the curve reaches half its plateau. For the Hill curve, set \(q=M_{\mathrm{ref}}/M_{\infty}\). A finite positive \(a_{50}\) requires \(0<q<1\).
The equation follows by multiplying the expression for \(q\) by its denominator: \(q(a_{\mathrm{ref}}^h+a_{50}^h)=a_{\mathrm{ref}}^h\). Rearranging gives \(q a_{50}^h=(1-q)a_{\mathrm{ref}}^h\). Dividing by \(q\), then taking the positive \(h\)th root, yields the displayed expression. Since \(q=M_{\mathrm{ref}}/M_{\infty}\), this converts a reference-age value to \(a_{50}\) when the plateau and Hill coefficient are specified.
If the plateau is fixed at 1, the reference value itself is \(q=F(a_{\mathrm{ref}})\). If the plateau is estimated, the ratio matters: for example, \(M_{\mathrm{ref}}=0.6\) and \(M_{\infty}=1.2\) give \(q=0.5\), not 0.6. Using the raw reference value as a fraction of the plateau would give the wrong half-maturation age.
Worked Example: Normalizing a Hill Curve at One Year
Suppose a hypothetical maturation fraction has \(a_{50}=1\) year and \(h=2\). Find its value at a reference age of 2 years, its normalized value at age 0.5 years, and the mature plateau of the reference-normalized multiplier.
At the reference age,
At age 0.5 years, \(F(0.5)=0.5^2/(0.5^2+1^2)=0.25/1.25=0.2000\). Therefore the multiplier relative to age 2 years is \(R(0.5;2)=0.2000/0.8000=0.2500\). At age 2 years, the ratio is 1 by definition. As age tends toward the mature plateau, \(F(a)\) tends to 1, so \(R(a;2)\) tends to \(1/0.8000=1.250\). The normalized multiplier is not bounded by 1.
Worked Example: Recovering Half-Maturation Age From a Reference Value
Suppose a hypothetical pathway has mature plateau \(M_{\infty}=1\), Hill coefficient \(h=2\), and maturation value \(M_{\mathrm{ref}}=0.60\) at \(a_{\mathrm{ref}}=1\) year. Calculate \(a_{50}\), then check the implied maturation value at age 2 years.
First, \(q=M_{\mathrm{ref}}/M_{\infty}=0.60/1=0.60\). Thus
The check at age 1 gives \(1/(1+0.8165^2)=1/(1+0.6667)=0.6000\), as required. At age 2,
This parameterization describes a value of 0.60 at one year and a mature plateau of 1. The value at age 2 is 0.8571, not 0.60; the latter is tied to the stated reference age.
Using a Reference Age in a Clearance Model
A common application is to model clearance using both body weight and maturation. For illustration, let \(WT\) be body weight, \(WT_{\mathrm{ref}}\) a fixed reference weight, and \(CL_{\mathrm{ref}}\) typical clearance at the reference weight and reference age. With an allometric exponent of 0.75 and a Hill maturation fraction, one form is
At \(a=a_{\mathrm{ref}}\) and \(WT=WT_{\mathrm{ref}}\), the model gives \(CL=CL_{\mathrm{ref}}\). The ratio ensures this anchoring. The units of \(CL_{\mathrm{ref}}\) are clearance units, such as L/h; the weight ratio and maturation ratio are dimensionless. This form assumes that the weight scaling and maturation ratio are multiplied as specified and that the chosen maturation curve is appropriate over the ages being modeled.
Worked Example: Interpreting a Clearance Reference
Consider a hypothetical model with \(CL_{\mathrm{ref}}=4.00\) L/h at \(WT_{\mathrm{ref}}=10\) kg and \(a_{\mathrm{ref}}=1\) year. Let \(a_{50}=2\) years and \(h=2\). Calculate typical clearance at age 3 years and weight 20 kg.
The maturation fractions are \(F(1)=1^2/(1^2+2^2)=1/5=0.2000\) and \(F(3)=3^2/(3^2+2^2)=9/13=0.6923\). Their ratio is \(0.6923/0.2000=3.462\). The allometric factor is \((20/10)^{0.75}=2^{0.75}=1.682\). Therefore,
The result is a model prediction at the stated age and weight. At the reference age and reference weight, the same equation returns 4.00 L/h. At a fixed weight, the maturation multiplier tends toward \(1/F(1)=5\) as age increases, because the plateau of \(F(a)\) is 1. It would be incorrect to interpret \(CL_{\mathrm{ref}}\) as the mature clearance at 10 kg.
Choosing and Changing the Reference
A useful reference age is one with a clear interpretation and relevant support in the application. A clinically meaningful age can make \(CL_{\mathrm{ref}}\) understandable; an age near the center of the observations can also make the anchor representative of the data. The choice does not establish a physiological transition by itself. It is a fixed point at which the model is expressed.
Changing the reference from \(a_{\mathrm{ref},1}\) to \(a_{\mathrm{ref},2}\) changes the numerical anchor. For the same underlying curve, the new value is \(M_{\mathrm{ref},2}=M_{\mathrm{ref},1}F(a_{\mathrm{ref},2})/F(a_{\mathrm{ref},1})\). In a normalized model, the new age multiplier must also use \(F(a)/F(a_{\mathrm{ref},2})\). Making both changes preserves predictions. Merely replacing the age in the denominator while keeping the same \(CL_{\mathrm{ref}}\) generally changes the curve’s scale.
| Parameterization | Direct interpretation | Important check |
|---|---|---|
| Plateau-based \(M_{\infty}F(a)\) | Value approaches \(M_{\infty}\) at mature ages | The age-specific value is below the plateau for finite ages under the Hill model |
| Reference value plus plateau | Value at \(a_{\mathrm{ref}}\) and value approached at maturity | For an increasing Hill curve, \(0<M_{\mathrm{ref}}<M_{\infty}\) |
| Reference-normalized ratio \(F(a)/F(a_{\mathrm{ref}})\) | Multiplier is 1 at \(a_{\mathrm{ref}}\) | The mature multiplier is \(1/F(a_{\mathrm{ref}})\), not 1 |
These are alternative expressions of the same curve only when their parameters are related consistently. In estimation, changing coordinates can affect numerical behavior, but it does not automatically add information or make a poorly supported curve well determined. If bounds or prior distributions are imposed, they must be translated to represent the same allowed curves; using identical numerical bounds on differently interpreted parameters can define a different model.
Assumptions, Checks, and Common Pitfalls
The conversions and interpretations above assume a positive Hill coefficient, a positive half-maturation age, a fixed reference age on the same age scale used throughout the model, and a positive reference age. The simple Hill curve has one increasing transition and approaches a single plateau. A reference-value parameterization also requires the reference value to be compatible with that curve: for an increasing curve, it must lie strictly between zero and the mature plateau. If the estimated reference value is at or above the plateau, the displayed conversion does not yield a finite positive \(a_{50}\).
Check that ages and \(a_{50}\) use the same units. If the model uses years, substituting an age recorded in months without converting it changes the curve. Check the fitted curve at the reference age, at younger and older ages, and in the limit toward the plateau. These checks can reveal an incorrect denominator, a reversed ratio, or an implausible interpretation. The reference age should also be reported with the parameter so a number such as “reference clearance” is not ambiguous.
- Pitfall: calling the reference-normalized multiplier a maturation fraction. Since its value at the reference age is 1, it may exceed 1 at older ages. Use \(F(a)\) for a fraction of the plateau and \(F(a)/F(a_{\mathrm{ref}})\) for a ratio to the reference value.
- Pitfall: changing the reference age but keeping the same anchor value. That generally changes predictions. Recalculate the anchor from the underlying curve, or refit with the new parameterization.
- Pitfall: confusing the reference value with the mature plateau. They are equal only if the reference age is at the plateau; for a finite age on a rising Hill curve, the reference value is below it.
- Pitfall: mixing age units. The ratio of ages in the Hill function is dimensionless only when both ages use the same units.
A plateau-based model is often easiest to interpret when the mature value is the main scientific quantity. A reference-age model is often convenient when a typical value at a defined age is the desired anchor. Neither form is universally preferable; the choice should be matched to the question, reported clearly, and implemented with parameter transformations that preserve the intended curve.
Check Your Understanding
Answer each question, then compare with the concise solutions below.
- A Hill fraction has \(a_{50}=3\) years and \(h=1\). What is its value at a reference age of 3 years, and what is the plateau of the multiplier normalized there?
- A curve has \(M_{\infty}=2.0\), \(M_{\mathrm{ref}}=1.0\), and \(a_{\mathrm{ref}}=4\) years. What is \(q\), and what is \(a_{50}\) if \(h=2\)?
- Why can a reference-normalized maturation multiplier be greater than 1 without contradicting its definition?
- At a fixed weight, a clearance model is normalized at age 2 years. What must be changed when the reference age is switched to 1 year to retain the same clearance predictions?
Solutions. (1) \(F(3)=3/(3+3)=0.5000\). The normalized multiplier’s plateau is \(1/0.5000=2.000\). (2) \(q=1.0/2.0=0.5000\), so \(a_{50}=4\sqrt{(1-0.5)/0.5}=4.00\) years. (3) It is a ratio to the value at the reference age, not a fraction of the mature plateau; later values can exceed the reference value. (4) Recalculate the reference clearance as the value of the same underlying curve at 1 year, and replace the maturation denominator with \(F(1)\). Keeping the old anchor while changing only the denominator would change predictions.