Quasi-Steady-State Reduction of the Full TMDD Model

Why Reduce the Full TMDD Model?

The full target-mediated drug disposition (TMDD) model represents free drug, free target, and drug–target complex as separate dynamic quantities. That detail can be important, but it also introduces a state variable for the complex whose formation and loss may occur much faster than the overall changes in drug and target. When complex adjustment is sufficiently fast, its concentration can be approximated algebraically while the slower drug and target quantities remain dynamic.

This reduction is called a quasi-steady-state (QSS) approximation. “Quasi” matters: the complex need not be at binding equilibrium, and the approximation does not assert that every state is constant. It says that the rate of change of complex is small compared with its formation and loss rates over the time interval where the approximation is used.

Definition: Let \(C\) be free drug concentration, \(R\) free target concentration, and \(B\) drug–target complex concentration. Under the QSS approximation, complex formation approximately balances total complex loss, so \(dB/dt\approx 0\). The resulting algebraic expression for \(B\) is substituted into the remaining dynamic equations.

Full Model and the QSS Derivation

Consider a well-mixed, one-to-one binding model with constant rate parameters. Free drug is eliminated at first-order rate \(k_{el}\); free target is synthesized at rate \(k_{syn}\) and lost at rate \(k_{deg}\); and complex is removed by internalization at rate \(k_{int}\). Let \(u(t)\) denote drug input expressed as a concentration rate. The mass-balance equations are

\begin{aligned} \frac{dC}{dt} &= u(t)-k_{el}C-k_{on}CR+k_{off}B,\\ \frac{dR}{dt} &= k_{syn}-k_{deg}R-k_{on}CR+k_{off}B,\\ \frac{dB}{dt} &= k_{on}CR-(k_{off}+k_{int})B. \end{aligned}

Here \(k_{on}\) has units of concentration\(^{-1}\) time\(^{-1}\), while \(k_{off}\), \(k_{int}\), \(k_{el}\), and \(k_{deg}\) have units of time\(^{-1}\). The synthesis rate has units of concentration per time. The model assumes that binding consumes one free drug and one free target, dissociation returns both, and internalization removes complex. These assumptions define the particular model being reduced; different binding stoichiometry or complex-loss pathways require different equations.

Set the complex formation rate approximately equal to the total complex loss rate:

k_{on}CR\approx(k_{off}+k_{int})B.

Define the QSS concentration constant

K_{SS}=\frac{k_{off}+k_{int}}{k_{on}}.

It has units of concentration. Rearranging the formation-loss balance gives \(B\approx CR/K_{SS}\). This expression contains free target \(R\); for a reduction using total target \(R_{tot}=R+B\), solve the same relation together with the total-target definition:

R_{tot}=R+B\approx R\left(1+\frac{C}{K_{SS}}\right).

Therefore,

R\approx\frac{R_{tot}K_{SS}}{K_{SS}+C}, \qquad B\approx\frac{R_{tot}C}{K_{SS}+C}.

The algebraic complex concentration is thus determined by free drug and total target. This derivation does not set \(dR/dt\) to zero or require total target to be constant.

The constant \(K_{SS}\) includes both dissociation and internalization. By contrast, the equilibrium dissociation constant is \(K_D=k_{off}/k_{on}\), which follows from balancing association with dissociation alone. Thus \(K_{SS}=K_D+k_{int}/k_{on}\). They coincide only when internalization is negligible relative to dissociation. Treating \(K_{SS}\) as \(K_D\) without checking that condition changes the predicted complex concentration.

Reduced Dynamic Equations

Adding the free-target and complex equations gives the total-target balance. Binding and dissociation cancel because they move target between the free and bound states:

\frac{dR_{tot}}{dt}=k_{syn}-k_{deg}R-k_{int}B.

Substitute the QSS expressions for \(R\) and \(B\) to obtain a dynamic equation for total target. For free drug, the full equation contains the net binding term \(-k_{on}CR+k_{off}B\). Under the formation-loss balance, \(k_{on}CR\approx(k_{off}+k_{int})B\), so

-k_{on}CR+k_{off}B\approx-k_{int}B.

The reduced system is consequently

\begin{aligned} \frac{dC}{dt} &\approx u(t)-k_{el}C-k_{int}\frac{R_{tot}C}{K_{SS}+C},\\ \frac{dR_{tot}}{dt} &\approx k_{syn} -k_{deg}\frac{R_{tot}K_{SS}}{K_{SS}+C} -k_{int}\frac{R_{tot}C}{K_{SS}+C}. \end{aligned}

The drug equation shows why complex internalization contributes to target-mediated drug loss: under QSS, association and dissociation nearly offset except for the part of complex loss caused by internalization. The reduction retains the changing free-drug and total-target concentrations; it removes \(B\) as an independently integrated state.

Worked Example: Calculating the QSS Complex Concentration

Suppose a hypothetical model has \(k_{on}=0.20\ \text{nM}^{-1}\text{h}^{-1}\), \(k_{off}=0.10\ \text{h}^{-1}\), and \(k_{int}=0.30\ \text{h}^{-1}\). At a particular time, let \(C=3.0\ \text{nM}\) and \(R_{tot}=10.0\ \text{nM}\). Calculate the QSS complex concentration and verify the formation-loss balance.

First, calculate the QSS constant:

K_{SS}=\frac{0.10+0.30}{0.20}=2.0\ \text{nM}.

Then,

B\approx\frac{R_{tot}C}{K_{SS}+C} =\frac{(10.0\ \text{nM})(3.0\ \text{nM})}{2.0\ \text{nM}+3.0\ \text{nM}} =6.0\ \text{nM}.

The implied free target is \(R=R_{tot}-B=10.0-6.0=4.0\ \text{nM}\). Association is \(k_{on}CR=(0.20)(3.0)(4.0)=2.4\ \text{nM/h}\). Total complex loss is \((k_{off}+k_{int})B=(0.10+0.30)(6.0)=2.4\ \text{nM/h}\), as required by the QSS balance.

For comparison, \(K_D=0.10/0.20=0.50\ \text{nM}\), not \(2.0\ \text{nM}\). The difference arises because internalization is appreciable in this hypothetical example; the QSS balance includes it as a complex-loss process.

When Is the Approximation Credible?

The QSS formula is not justified merely because it can be written down. It is most defensible when complex adjusts faster than the quantities driving its formation and loss, particularly \(C\) and \(R_{tot}\). If these quantities change only gradually over the complex-adjustment timescale, the complex can remain close to its instantaneous QSS value after any initial rapid transient.

A useful local timescale check follows by temporarily holding \(C\) and \(R_{tot}\) fixed. Since \(R=R_{tot}-B\), the complex equation becomes

\frac{dB}{dt}=k_{on}C R_{tot} -\left(k_{on}C+k_{off}+k_{int}\right)B.

A small perturbation from the corresponding algebraic value decays at rate \(k_{on}C+k_{off}+k_{int}\), giving the local relaxation time

\tau_B=\frac{1}{k_{on}C+k_{off}+k_{int}}.

This is a local diagnostic, not a universal guarantee: in the full model, \(C\) and \(R_{tot}\) also move. Compare \(\tau_B\) with the timescale of their changes. A substantially shorter \(\tau_B\) supports QSS; similar or longer timescales warn that the complex may lag behind its algebraic value. Initial conditions also matter: if the initial complex concentration is far from the QSS value, the approximation may be poor during the initial adjustment even when later behavior is well approximated.

Worked Example: Comparing Complex and Drug Timescales

Use \(k_{on}=0.20\ \text{nM}^{-1}\text{h}^{-1}\), \(k_{off}=0.10\ \text{h}^{-1}\), \(k_{int}=0.30\ \text{h}^{-1}\), and \(C=3.0\ \text{nM}\). At this concentration, the relaxation rate is

k_{on}C+k_{off}+k_{int}=(0.20)(3.0)+0.10+0.30=1.0\ \text{h}^{-1},

so \(\tau_B=1.0\ \text{h}\). Suppose, hypothetically, that free drug is declining at \(0.08\ \text{nM/h}\) and total target changes little over the next hour. For fixed \(R_{tot}=10.0\ \text{nM}\), the algebraic expression is \(B_{SS}=R_{tot}C/(K_{SS}+C)\). Its local sensitivity to \(C\) is

\frac{dB_{SS}}{dC} =\frac{R_{tot}K_{SS}}{(K_{SS}+C)^2} =\frac{(10.0)(2.0)}{(2.0+3.0)^2} =0.80.

Thus the instantaneous algebraic complex value would change at about \((0.80)(-0.08)=-0.064\ \text{nM/h}\). The changing algebraic value is modest relative to the \(2.4\ \text{nM/h}\) association and loss fluxes calculated in the previous example, supporting—but not proving—the approximation under these slow-change conditions.

If instead free drug changed at \(12\ \text{nM/h}\) near the same concentration, the algebraic complex value would change at about \((0.80)(-12)=-9.6\ \text{nM/h}\). That forcing is large relative to the complex fluxes, and the one-hour relaxation time is not short compared with the rapid drug change. QSS would be questionable in that interval; integrating the complex dynamically is safer.

Using and Checking the Reduced Model

Worked Example: Calculating Reduced-Model Rates

Continue with the hypothetical values \(C=3.0\ \text{nM}\), \(R_{tot}=10.0\ \text{nM}\), and \(B\approx6.0\ \text{nM}\), so \(R\approx4.0\ \text{nM}\). Let \(u(t)=2.0\ \text{nM/h}\), \(k_{el}=0.10\ \text{h}^{-1}\), \(k_{syn}=1.0\ \text{nM/h}\), and \(k_{deg}=0.20\ \text{h}^{-1}\). Compute the reduced instantaneous rates.

For free drug,

\frac{dC}{dt}\approx 2.0-(0.10)(3.0)-(0.30)(6.0) =2.0-0.30-1.80 =-0.10\ \text{nM/h}.

For total target,

\frac{dR_{tot}}{dt}\approx 1.0-(0.20)(4.0)-(0.30)(6.0) =1.0-0.80-1.80 =-1.60\ \text{nM/h}.

These are instantaneous derivatives at the stated concentrations, not predictions that the same rates persist. As \(C\) and \(R_{tot}\) change, the algebraic estimate of \(B\) and therefore both derivatives change as well.

A practical assessment should examine whether the approximation is adequate for the intended use, not just whether its timescale argument sounds plausible. Useful checks include comparing full and reduced model predictions over the observed dosing and sampling intervals, inspecting discrepancies around rapid input changes, and checking whether estimated parameters or predictions are sensitive to retaining the dynamic complex state. If the full and reduced models disagree materially in the time window that informs the analysis, the reduced model may be too restrictive.

The main alternative is to retain the full TMDD equations and integrate \(C\), \(R\), and \(B\) dynamically. This avoids the QSS assumption but requires an additional state and may involve faster dynamics. An equilibrium-based algebraic treatment is a different approximation: it balances association with dissociation rather than with dissociation plus internalization. It is not interchangeable with QSS when internalization is appreciable. Model choice depends on the processes represented and on whether the approximation reproduces the concentrations and time course needed for the scientific question.

Common Mistakes and Limitations

Calling QSS binding equilibrium. QSS balances association against dissociation and internalization. The derivation does not require microscopic reversibility or negligible internalization.
Assuming the complex is constant. QSS means its derivative is approximately zero relative to the processes that determine it; its algebraic value can still change as \(C\) or \(R_{tot}\) changes.
Ignoring the initial transient. If the initial \(B\) differs substantially from its QSS value, an early interval may not satisfy the approximation. A full model may be needed for that interval.
Using the reduction during abrupt changes without checking it. Rapid bolus input, rapid concentration decline, or abrupt changes in target can outpace complex adjustment.
Assuming a good fit proves the mechanism. A reduced model can describe observed data while misrepresenting unobserved complex dynamics. Its adequacy is conditional on the data, design, and intended predictions.

The reduction also inherits the structural assumptions of the full model: one-to-one binding, specified loss pathways, and well-mixed concentrations. It does not correct misspecified binding or target processes. Sparse measurements may make full and reduced models difficult to distinguish, so predictive agreement over observed data should not be treated as proof that the QSS condition holds at every time.

Key takeaway: QSS follows by balancing complex association with dissociation plus internalization, giving \(B\approx R_{tot}C/(K_{SS}+C)\) with \(K_{SS}=(k_{off}+k_{int})/k_{on}\). Its validity depends on complex adjustment being fast relative to changes in free drug and total target, not on binding equilibrium.

Check Your Understanding

Answer each question using the model and assumptions stated above.

If \(k_{off}=0.06\ \text{h}^{-1}\), \(k_{int}=0.14\ \text{h}^{-1}\), and \(k_{on}=0.10\ \text{nM}^{-1}\text{h}^{-1}\), calculate \(K_{SS}\).
For \(K_{SS}=2.0\ \text{nM}\), \(C=2.0\ \text{nM}\), and \(R_{tot}=8.0\ \text{nM}\), calculate \(B\) and \(R\) under QSS.
Why does \(K_{SS}\) generally differ from \(K_D\) when complex internalization occurs?
State one condition that makes QSS less credible and one diagnostic comparison that can help assess it.

Solutions. (1) \(K_{SS}=(0.06+0.14)/0.10=2.0\ \text{nM}\). (2) \(B\approx(8.0)(2.0)/(2.0+2.0)=4.0\ \text{nM}\), and \(R=R_{tot}-B=4.0\ \text{nM}\). (3) \(K_{SS}=(k_{off}+k_{int})/k_{on}\), whereas \(K_D=k_{off}/k_{on}\); internalization adds to total complex loss. They are equal only if \(k_{int}=0\), or approximately equal when internalization is negligible relative to dissociation. (4) Rapid changes in drug or target relative to complex relaxation make QSS less credible. Compare full and reduced model predictions over the time window of interest, especially around rapid changes.