Dynamics of targetmediated drug disposition: characteristic profiles and parameter identification
 1.7k Downloads
 36 Citations
Abstract
In this paper we present a mathematical analysis of the basic model for target mediated drug disposition (TMDD). Assuming high affinity of ligand to target, we give a qualitative characterisation of ligand versus time graphs for different dosing regimes and derive accurate analytic approximations of different phases in the temporal behaviour of the system. These approximations are used to estimate model parameters, give analytical approximations of such quantities as area under the ligand curve and clearance. We formulate conditions under which a suitably chosen Michaelis–Menten model provides a good approximation of the full TMDDmodel over a specified time interval.
Keywords
Target Receptor Antibodies Drugdisposition Michaelis–Menten Quasisteadystate Quasiequilibrium Singular perturbationIntroduction
The interaction of ligand and target in the process of drugdisposition offers interesting examples of complex dynamics when target is synthesised and degrades and when both ligand and ligand–target complex are eliminated. In recent years such dynamics has received considerable attention because it is important in the context of data analysis, but also, more generally, in the context of system biology because this model serves as a module in more complex systems [1].
Based on conceptual ideas developed by Levy [2], the basic model for target mediated drug disposition (TMDD) was formulated by Mager and Jusko [3]. Earlier studies of ligand–target interactions go back to Michaelis and Menten [4]. We also mention ideas about receptor turnover developed by Sugiyama and Hanano [5]. Mager and Krzyzanski [6] showed how rapid binding of ligand to target leads to a simpler model, Gibiansky et al. [7] studied the related quasisteadystate approximation to the model and Marathe et al. [8] conducted a numerical validation of the rapid binding approximation. Gibiansky et al. [9] also pointed out a relation with the classical indirect response model. For further background we refer to the books by Meibohm [10] and Crommelin et al. [11], and to the reviews by Lobo et al. [12] and Mager [13].
In practice, the Michaelis–Menten model is often used when ligand curves exhibit TMDD characteristics (see e.g. Bauer et al. [14]). Recently, Yan et al. [15] analysed the relationship between TMDD and Michaelis–Menten type dynamics. We also mention the work by Krippendorff et al. [16] which studies an extended TMDD system which includes receptor trafficking in the cell.
The characteristic features of TMDD dynamics were first studied in [3] under the condition of a constant target pool, i.e., the total amount of target: free and bound, was assumed to be constant in time. Under the same assumption, a mathematical analysis of this model was offered by Peletier and Gabrielsson [17]. This assumption was made, in part for educational reasons, because it makes a transparent geometric description possible, in which qualitative and quantitative properties of the dynamics can be identified and illustrated. In a recent paper Ma [18] compared different approximate models under the same assumption of a constant target pool.
In the present paper we extend this analysis to the full TMDD model and do not make the assumption that the target pool is constant. This means, in particular, that we shall now also be enquiring as to how the target pool changes over time and how it is affected by the dynamics of its zeroth order synthesis and first order degeneration.
 1.Properties of concentration profiles When the initial ligand concentration is larger than the endogenous receptor concentration, the dynamics of targetmediated drug disposition results in a characteristic ligand versus time profile. In Fig. 1 we show such a profile schematically.
One can distinguish four different phases in the dynamics of the system in which different processes are dominant: (A) a brief initial phase, (B) an apparent linear phase, (C) a transition phase and (D) a linear terminal phase. We obtain precise estimates for the duration of each of these phases and for each of them we obtain accurate analytical estimates for the concentration versus time graphs of the ligand, the receptor and the ligand–receptor complex.
On the basis of ligand concentration versus time curves we will develop instruments for extracting information about the target and the ligand–target complex versus time curves.
 2.
Parameter identifiability We shed light on what we can predict when we have only measured (a) the ligand, (b) ligand and target, (c) ligand and complex, and (d) all of the above.
 3.
Systems analysis Whilst focussing on concentration versus time curves we gain considerable qualitative understanding and quantitative estimates about the impact of the different parameters in the model and on quantities such as the area under the curve and time to steady state of the different compounds.
 4.
Model comparison An important issue is the question as to how the full TMDD model compares with the simpler Michaelis–Menten model [7, 8, 15]. In this paper we point out how the full model and the reduced Michaelis–Menten model differ significantly in the initial second order phase and in the linear terminal phase, in that the terminal rate (λ_{ z }) of ligand in the full model is much smaller than that in the Michaelis–Menten model.

(i) Through a bolus dose. Then k _{ f } = 0. We denote the initial ligand concentration by L(0) = L _{0} = D/V _{ c }, where D is the dose and V _{ c } the volume of the central compartment.

(ii) Through a constant rate infusion In _{ L }. Then k _{ f } = In _{ L }/V _{ c } > 0 and L(0) = 0.
Anchoring our investigation in a case study in which ligand is administered through a series of bolus doses, we dissect the resulting time courses of the three compounds, L, R and RL and identify characteristic phases, Phases A–D shown in Fig. 1. We associate these phases with specific processes and show, using singular perturbation theory [19, 20, 21], that individual phases may be analysed through appropriately chosen simplified models, yielding accurate closedform approximations. They offer tools which may be used to compute critical quantities such as residence time, and to verify whether different approximations to the full TMDD model, such as the rapid binding approximation [6] and the quasisteadystate approximation [7, 18] are valid in the different phases. These issues are discussed at the conclusion of this paper.
Much of the mathematical analysis underpinning the results presented throughout the text is presented in a series of Appendices at the end of the paper.
Case study
Preselected parameter values
Symbol  Unit  Value 

V _{ t }  L/kg  0.1 
Cl _{ d }  (L/kg)/h  0.003 
Cl _{(L)}  (L/kg)/h  0.001 
k _{on}  (mg/L)^{−1}/h  0.091 
k _{off}  1/h  0.001 
k _{in}  (mg/L)/h  0.11 
k _{out}  1/h  0.0089 
k _{ e(RL)}  1/h  0.003 
R _{0}  mg/L  12 
Data analysis
The purpose of this study is to demonstrate the possibility of fitting the eightparameter model shown in Fig. 2, to three different sets of high quality data with increasing richness, and show how precision of the estimates of the model parameters increases when successively information about target (II) and target and complex (III) is added. We use this data set for two purposes: (i) for data analysis and (ii) for highlighting critical features of the temporal behaviour of the three compounds.
Simulated data from three sources (ligand, target and complex) were intentionally used. We have experienced that data of less quality gave biased and imprecise estimates as well as biased and imprecise predictions of ligand, target and complex.
Final parameter estimates and their relative standard deviation (CV%) on the basis of the three datasets
Symbol  Unit  I (L)  II (L & R)  III (L & R & RL) 

V _{ t }  L/kg  0.101 (2)  0.100 (2)  0.100 (1) 
Cl _{ d }  (L/kg)/h  0.003 (4)  0.003 (3)  0.003 (3) 
Cl _{(L)}  (L/kg)/h  0.001 (1)  0.001 (1)  0.001 (1) 
k _{on}  (mg/L)^{−1}/h  0.099 (17)  0.092 (2)  0.096 (1) 
k _{off}  1/h  0.001 (27)  0.001 (13)  0.001 (3) 
k _{out}  1/h  0.009 (6)  0.009 (2)  0.009 (2) 
k _{ e(RL)}  1/h  0.002 (27)  0.002 (23)  0.002 (2) 
R _{0}  mg/L  12 (4)  12 (1)  12 (1) 
Dataset I is made up from simulated concentration–time profiles covering five orders of magnitude in concentration range and from 0 to 500 h. Dataset II contains the same simulated ligand (L) profiles as in dataset I as well as target (R) concentration–time profiles obtained at each dose level. Dataset III includes dataset II but is enriched by four simulated timecourses of the ligand–target complex (LR) as well.
The four doses are D = 1.5, 5, 15 and 45 mg/kg. The volume V _{ c } of the central compartment being 0.05 L/kg, this yields the following initial ligand concentrations L _{0} = D/V _{ c } = 30, 100, 300 and 900 mg/L.

Dataset I—which involves L—allows the prediction of robust ligand concentration–time profiles within the suggested concentration and time frame. We see that if only ligand data are available, the majority of parameters except for k _{on}, k _{off} and k _{ e(RL)} are estimated with high precision. The latter three parameters are still highly dependent on information about the time courses of either target and/or complex. Since k _{kon}, k _{off} and k _{ e(RL)} have low precision (high CV%) we would discourage the use of these parameters for the prediction of tentative target and complex concentrations.

Dataset II—which involves L and R—still gives good precision in all parameters except k _{off} and k _{ e(RL)}, which will also be highly correlated. Since we also have experimental data of the target we encourage the use of this model for interpolation of target concentration–time courses, but not for concentration–time courses of the complex.

Dataset III—which includes L and R, as well as RL—gives high precision in all parameters. Since we also have measured the complex concentration–time course with high precision we obtained k _{off} and k _{ e(RL)} values with high precision. We doubt that the practical experimental situation can get very much better than this latter case where we have simultaneous concentration–time courses of L, R and RL with little experimental error due to biology and bioanalytical methods. Dataset III is an ideal case; the true experimental situation seldom gets better.
We also doubt the practical value of regressing too elaborate models to data. Models that capture the overall trend nicely but result in parameters with low precision and biased estimates may be of little value.
The volume of the central compartment V _{ c } ought to fall somewhere in the neighbourhood of the plasma water volume (0.05 L/kg) for large molecules in general and antibodies in particular. In our own experience of antibody projects this has been the case when data contained an acceptable granularity within the first couple of hours after the injection of the test compound. Therefore we assumed V _{ c } to be a constant term (0.05 L/kg) in this analysis and not part of the list of parameters to be estimated. We think this increases the robustness of the estimation procedure and is biologically viable.
Critical features of the graphs
 (a)
Initially all the ligand graphs in Fig. 3 exhibit a rapid drop which increases in relative sense as the ligand dose decreases. Over this initial period, which we refer to as Phase A, (cf. Fig. 1), R(t) exhibits a steep drop that becomes deeper as the drug dose increases.
 (b)
After the brief initial adjustment period, the graphs for large doses reveal linear first order kinetics over a period of time (Phase B) that shrinks as the drug dose decreases. At the lowest dose the linear period has vanished and the graph exhibits nonlinear kinetics.
 (c)
For the larger doses, there is an upward shift of the linear phase that appears to be linearly related to the ligand dose; the slope of this linear phase appears to be doseindependent.
 (d)
The point of inflection in the log(L) versus time curve—the middle of Phase C—which we observe in the graphs for L _{0} = 100 and 300, moves to the right as the initial dose increases, but stays at the same level. This is clearly seen in Fig. 3 in which the baseline value R _{0} and the value of K _{ d } and K _{ m } are also shown.
 (e)
For the lower doses we see that the log(L) versus time curve eventually becomes linear again, with a slope that is markedly smaller than it was in the nonlinear Phase C that preceded it. This part of the graph corresponds to Phase D in Fig. 1.
Summarising, in the ligand graphs of Fig. 3 we see for the higher drug doses the different phases A–D that were pointed out in Fig. 1. In the following analysis we explain these features and quantify them in that, for instance, we present estimates for the upward shift referred to in (c) and the rightward shift of the inflection point alluded to in (d).
Dynamics after a bolus administration
Simulations
Parameter estimates used for demonstrating the dynamics of L, R and RL after bolus and constantrate infusion regimens of L
Symbol  Unit  Value 

k _{ e(L)}  1/h  0.0015 
k _{on}  (mg/L)^{−1}/h  0.091 
k _{off}  1/h  0.001 
k _{in}  (mg/L)/h  0.11 
k _{out}  1/h  0.0089 
k _{ e(RL)}  1/h  0.003 
R _{0}  mg/L  12 
 (a)
A rapid initial adjustment (see the blowup on the right).
 (b)
A first linear phase with a slope which is independent of the dose, and which shifts upwards as the drug dose increases.
 (c)
A transition phase which shifts to the right as the drug dose increases, but maintains its level.
 (d)
A final linear terminal phase with a slope λ_{ z } that is again independent of the drug dose. For the parameter values of Table 3 we find that λ_{ z } ≈ k _{ e(RL)} = 0.003.
Alternatively, if k _{ e(RL)} > k _{out}, then R _{*} < R _{0} and we show that the total target pool first decreases before it returns to the baseline value R _{0}.

(ii) As the drug dose increases, R(t) ≈ 0 for an increasing time interval and the graphs of RL(t) and R _{tot}(t) trace—for the same increasing time interval—a common curve \(\Upgamma\) in the (t, R _{tot})plane (cf. Fig. 9). This curve \(\Upgamma\) is monotonically increasing and tends to the limit R _{*} as \(t \to \infty. \) If k _{ e(RL)} > k _{out}, we show that an analogous phenomenon occurs along a curve \(\Upgamma, \) which still tends to R _{*}, but is now decreasing.

(i) The three graphs exhibit a kink (a sharp angle) which shifts to the right (increasing time) as the drug dose increases.

(ii) R(t) − R _{0} tends to zero as \(t \to \infty\) in a biexponential manner, whilst RL(t) and R _{tot}(t) − R _{0} converge to zero in a monoexponential way.
Low dose graphs
We conclude these simulations with a comparison of highdose and lowdose graphs. We do this by adding simulations for initial ligand concentrations which are smaller than R _{0}. Specifically, we add the values L _{0} = 0.3, 1, 3 and 10 mg/L to the graphs shown in Figs. 5 and 6.
Mathematical analysis
Phase A
Phase B
Phase C
Phase D
When L(t) ≪ K _{ d }, i.e., beyond T _{3}, the ligand concentration is so small that the dynamics is linear again.
The critical times T _{1}, T _{2} and T _{3} provide a natural division of the dynamics in four phases: A, B, C and D, as was done in Fig. 1. In Phase A (0 < t < T _{1}), ligand, receptor and complex reach quasiequilibrium, in Phase B (T _{1} < t < T _{2}), the bulk of the ligand is eliminated from the system while most of the receptor is bound to ligand and in quasiequilibrium. Phase C (T _{2} < t < T _{3}) is a nonlinear transitional phase in which L exhibits a steep drop, and finally, in Phase D (\(T_3<t<\infty\)) the three compounds converge linearly towards their baseline values.
Receptor graphs: Phase B
In Fig. 9 we see how for the different drug doses, the simulations of R _{tot}(t) follow the graph \(\Upgamma\) up till some time, when they suddenly depart from \(\Upgamma. \)
Remark
We recall from the approximation (17) that R(t) ≈ 0 for T _{1} < t < T _{2} and hence that RL(t) ≈ R _{tot} in this phase of the dynamics, as we see confirmed in Fig. 6.
Ligand graphs: Phase B
Remark

(i) the ratio μ of R _{0} and L _{0} and

(ii) the ratio κ of the direct elimination rates of receptor, k _{out}, and ligand, k _{ e(L)}.
Ligand and receptor graphs: Phases C and D
In the simulations shown in Figs. 5 and 6 we see that in Phase C, L drops rapidly from O(10 × K _{ d }) to O(0.1 × K _{ d }), i.e., by a factor 100, whilst R _{tot} stays relatively close to R _{*} and changes by no more than a factor 1/7 ≈ 0.15. This suggests making the following assumption:
Assumption
R _{tot}(t) ≈ R _{*} or v(t) ≈ R _{*}/R _{0} in Phase C.

(i) the terminal slope λ _{ z } ^{TMDD} (k _{ e(RL)});

(ii) the intercept of the asymptote of log{L(t)} of the ligand graph in the terminal Phase D with the vertical line {t = T _{2}}.
For completeness we also compute the terminal slope by means of a standard analysis of the full TMDD system. This is done in Appendix 7. It is found that for the parameter values in Table 3, the terminal slope λ_{ z } of all the compounds, is given—to good approximation—by λ _{ z } ^{TMDD} = k _{ e(RL)}. This confirms the limit in (36) and the exponent in (38).
Comparison with Michaelis–Menten kinetics
Thus, the TMDDmodel and the Michaelis–Menten (MM)model exhibit very different terminal slopes, unless one also includes a nonspecific peripheral volume distribution term in the MMmodel.
The reduced model mimics the concentration–time data for the two highest doses reasonably well, whereas the two lower doses display systematic deviations between observed and predicted data.
Since the reduced model has two parallel elimination pathways (linear and nonlinear) it has the intrinsic capacity of exhibiting linear firstorder kinetics at low and at high concentrations. In the concentrationrange in between it behaves nonlinearly. For higher concentrations the MMroute is saturated and the linear elimination pathway dominates so that the system behaves linearly.
However, the typical concentration–time pattern for ligand seen in a true TMDD system (cf. Figs. 5, 8, 10), cannot be fully described by the parallel linear and MMelimination model. The reduced model displays typical biexponential decline (which is expected from a twocompartment model) at lower concentrations. That is generally not the case with the full TMDD model.
A clear distinction of the two models occurs initially, immediately after dosing (Phase A), when the secondorder reaction between ligand and circulating target forms the complex. This process cannot be captured by the reduced model, which may cause biased estimates (too large) of the central volume.
Symbol  Unit  Value  CV% 

V _{ c }  L/kg  0.05  – 
V _{ t }  L/kg  0.1  10 
Cl _{ d }  (L/kg)/h  0.00307  20 
Cl _{(L)}  (L/kg)/h  0.00090  10 
V _{max}  mg/h  0.0146  40 
K _{ M }  mg/L  3.68  50 
Phases that can be explained by the two MMmodels and the TMDDmodel
Phase  MMmodel (41)  MMmodel (45)  TMDDmodel (9) 

A  −  −  + 
B  +  +  + 
C  +/−  +  + 
D  −  −  + 
In this table, a plus (+) means that the corresponding phase can be adequately explained, whilst a minus (−) means that it cannot.
Constant rate drug infusion
When the infusion lasts long enough, i.e., when t _{washout} is large enough, the concentrations will converge towards their steady state values L _{ss}, R _{ss} and RL _{ss}. Then, at washout, they will return to their preinfusion values: L = 0, R = R _{0} and RL = 0.
Steady state concentrations of L, R and RL
The formula (48) for the ligand shows that L _{ss} will be smaller than expected from the ratio of ligand infusion ratetoclearance (In _{ L }/Cl _{(L)} = k _{ f }/k _{ e(L)}), due to the removal of ligand as part of the complex RL _{ss}. The same reasoning may be used to explain why the circulating target concentration R _{ss} given by (49) is smaller than the baseline concentration R _{0} = k _{in}/k _{out}. Due to the removal of target by means of the complex, the target concentration R _{ss} will drop further as the infusion rate increases and RL _{ss} increases accordingly (cf. Eq. (49)).
Using the parameter values in Table 3, we obtain 1/k _{ e(L)} = 667, k _{in}/k _{ e(L)} = 73, R _{*} = k _{in}/k _{ e(RL)} = 36.7 and log(K _{ m }) + log(R _{*}) = 0.45. We see that these values are confirmed by the numerically obtained graphs shown in Fig. 12.
We note that the expressions (47)–(49) show that the steady state concentrations do not depend on the on and off rates k _{on} and k _{off} individually, but only as part of the constant K _{ m }.
Simulations
The washout dynamics is very similar to the dynamics after a bolus dose, as described before: Phase A (0,T _{1}) now covers the infusion period so that T _{1} coincides here with the time of washout. Phases B–D are plainly evident in the postwashout dynamics.
The dynamics of the receptor R and the receptor–ligand complex RL are shown in Figs. 14 and 15. As in Phase A in the bolus administration, the predose receptor pool (R _{0}) quickly binds to the ligand. We see that the speed of receptor depletion increases with increasing infusion rate k _{ f }, consistent with the halflife estimate (15) after a bolus administration.
After washout, when k _{ f } is large, R(t) ≈ 0 and RL(t) ≈ R _{*} for a while before they abruptly return to their baseline values. It is evident from Fig. 15 that, as in Fig. 7, initially the slope of log(R _{0} − R) is steeper than that of log(RL). This is in agreement with the analysis presented in Appendix 5, where it is shown that over that period the halflives of R(t) − R _{0} and RL(t) are, respectively, O(1/k _{out}) and O(1/k _{ e(RL }).
Discussion and conclusion
When ligand is administered through a bolus dose, L _{0} > R _{0}, and the conditions in (58) are satisfied, four phases can be distinguished in the ligand elimination graph: a brief initial Phase A, a slow linear Phase B, a rapid nonlinear Phase C and then again a slow linear terminal Phase D (cf. Fig. 1). Thanks to accurate analytical approximations for these four phases as shown in the Eqs. (14) for Phase A, (20), (25)–(27) for Phase B, (36) for Phase C and (38) and (40) for Phase D, we may extract information about the model parameters.
Information contained in the four phases
Phase  TMDDmodel (9) 

A  R _{0} and k _{on} 
B  k _{ e(L)}, k _{in} (k _{out}) 
C  K _{ d } (k _{off}) 
D  k _{ e(RL)} (K _{ m }) 
It should be noted though that estimating R _{0} may be difficult, since often there are no data for the first phase because it is over very quickly.
The four phases identified in the ligand elimination graph after a bolus administration are reflected in the structure of the receptor versus time graphs (receptor, receptor–ligand complex, and total amount of receptor). During Phase A the receptor pool is quickly depleted, and it remains so during Phase B. Then, during the Phases C and D it climbs back to the terminal baseline level R _{0}.
 1.The rapid binding model [6, 18], in which it is assumed thatwhere K _{ d } is defined in (7). In Phase B, which is characterised by R(t) ≈ 0, we have dR/dt ≈ 0 so that, according to the second equation of the system (9), we have approximately$$ L \cdot R = K_d RL $$(59)which disagrees with (59).$$ \Updelta \mathop{=}\limits^{{\rm def}} L \cdot R  K_d RL =\frac{k_{\rm in}}{k_{\rm on}} $$(60)
In contrast, in Phases C and D the identity (59) is satisfied according to the results established in Appendix 5 (cf. (102)), and Appendix 7 where it was shown that λ_{ z } = k _{ e(RL)}.
In Fig. 17 we show how the quantity \(\Updelta = L \cdot R  K_d RL\) varies with time and how \(\Updelta\) rapidly jumps from k _{in}/k _{on} down to zero at the transition of Phases B and C.
 2.The quasisteadystate model [7, 18] in which it is assumed thatwhere K _{ m } is defined in (8). Evidently, this assumption is not valid during Phases C and D, but it is during that part of Phase B in which RL(t) ≈ R _{*}. In that interval dRL/dt ≈ 0 (cf. Fig. 6) and hence, by the third equation of the system (1), condition (61) is approximately satisfied.$$ L \cdot R = K_m RL $$(61)
Anchored on the data of the Case Study, the analysis in this paper is based on the Assumptions A, B and C (or C^{*}). The question arises whether the characteristic features of the ligand elimination curves, such as shown in Fig. 1, are still present when these assumptions are not met.
In general, the behaviour of nonlinear systems such as (1) is very sensitive to the values of the parameters and initial data involved. However, a number of features of the ligand versus time graphs is quite robust in that they may survive if e.g. Assumption A is not satisfied and K _{ d } and R _{0} are comparable. Thus, the estimate (15) for T _{1} suggests that the initial Phase A will remain short relative to typical times over which the other processes develop when K _{ d }/L _{0} is small. We refer to [17] and [18] for a detailed analysis of this situation.
The approximate expressions for L and R _{tot} in Phase B (Eqs. (25) and (20)) are still valid provided that R(t) ≈ 0. This will still be the case when Assumption A is replaced by K _{ d } ≪ L _{0} [17, 18].
In contrast, the analysis of the dynamics in Phase C that is carried out in Appendix 5 depends critically on Assumption A. It will be interesting to study the dynamics beyond Phase B when Assumption A does not hold, as it will be interesting to see how the value of α, β and γ affects the dynamics.
We have selected a set of data (ligand and circulating target and complex) with low experimental variability, concentration–time courses at four ligand doses given as bolus injections, and wellspaced data in time that captures the necessary phases and shapes of a typical TMDD system. Based on this approach and the mathematical/analytical analysis, we can draw conclusions about the identifiability of the model parameters and appropriate system. When data are less precise and information rich, or, when target and/or complex are less accessible, the a priori expectations of parameter accuracy and precision will be lower.
Footnotes
Notes
Acknowledgments
The constructive criticisms of the manuscript of the reviewers improved the quality of the text, and are highly appreciated.
Open Access
This article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution, and reproduction in any medium, provided the original author(s) and the source are credited.
References
 1.Shankaran H, Rasat H, Wiley HS (2007) Cell surface receptors for signal transduction and ligand transport: a design principles study. PloS Comput Biol 3(6):986–999CrossRefGoogle Scholar
 2.Levy G (1994) Pharmacologic target mediated drug disposition. Clin Pharmacol Ther 56:248–252PubMedCrossRefGoogle Scholar
 3.Mager DE, Jusko WJ (2001) General pharmacokinetic model for drugs exhibiting targetmediated drug disposition. J Pharmacokinet Phamacodyn 28(6):507–532CrossRefGoogle Scholar
 4.Michaelis L, Menten ML (1913) Die Kinetik der Invertinwirkung. Biochem Z 49:333–369Google Scholar
 5.Sugiyama Y, Hanano M (1989) Receptormediated transport of peptide hormones and its importance in the overall hormone disposition in the body. Pharm Res 6(3):192–202PubMedCrossRefGoogle Scholar
 6.Mager DE, Krzyzanski W (2005) Quasiequilibrium pharmacokinetic model for drugs exhibiting targetmediated drug disposition. Pharm Res 22(10):1589–1596PubMedCrossRefGoogle Scholar
 7.Gibiansky L, Gibiansky E, Kakkar T, Ma P (2008) Approximations of the targetmediated drug disposition model and identifiability of model parameters. J Pharmacokinet Pharmacodyn 35(5):573–591PubMedCrossRefGoogle Scholar
 8.Marathe A, Krzyzanski W, Mager D (2009) Numerical validation of a rapid binding approximation of a targetmediated drug disposition pharmacokinetic model. J Pharmacokinet Pharmacodyn 36:199–219PubMedCrossRefGoogle Scholar
 9.Gibiansky L, Gibiansky E (2009) Targetmediated drug disposition model: relationships with indirect response models and application to population PKPD analysis. J Pharmacokinet Pharmacodyn 36:341–351PubMedCrossRefGoogle Scholar
 10.Meibohm, B (eds) (2006) Pharmacokinetics and pharmacodynamics of biotech drugs: principles and case studies in drug development. WileyVCH Verlag GmbH & Co KGaA, WeinheimGoogle Scholar
 11.Crommelin DJA, Sindelar RD, Meibohm B (eds) (2008) Pharmaceutical biotechnology: fundamentals and applications. Informa Healthcare, New YorkGoogle Scholar
 12.Lobo E, Hansen RJ, Balthasar JR (2004) Antibody pharmacokinetics and pharmacodynamics. J Pharm Sci 93(11):2645–2667PubMedCrossRefGoogle Scholar
 13.Mager D (2006) Targetmediated drug disposition and dynamics. Biochem Pharmacol 72:1–10PubMedCrossRefGoogle Scholar
 14.Bauer RJ, Dedrick RL, White ML, Murray MJ, Garovoy MR (1999) Population pharmacokinetics and pharmacodynamics of the antiCD11a antibody hu1124 in human subjects with psoriasis. J Pharmacokinet Biopharm 27:397–420PubMedGoogle Scholar
 15.Yan X, Mager DE, Krzyzanski W (2010) Selection between Michaelis Menten and targetmediated drug disposition pharmacokinetic models. J Pharmacokinet Pharmacodyn 37:25–47PubMedCrossRefGoogle Scholar
 16.Krippendorff BF, Kuester K, Kloft C, Huisinga W (2009) Nonlinear pharmacokinetics of therapeutic proteins resulting from receptor mediated endocytosis. J Pharmacokinet Pharmacodyn 36:239–260PubMedCrossRefGoogle Scholar
 17.Peletier LA, Gabrielsson J (2009) Dynamics of targetmediated drug disposition. Eur J Pharm Sci 38:445–464PubMedCrossRefGoogle Scholar
 18.Ma P (2012) Theoretical considerations of targetmediated drug disposition models: simplifications and approximations. Pharm Res 29:866–882PubMedCrossRefGoogle Scholar
 19.Segel LA (1988) On the validity of the steady state assumption of enzyme kinetics. Bull Math Biol 50(6):579–593PubMedGoogle Scholar
 20.Segel LA, Slemrod M (1989) The quasisteady state assumption: a case study in perturbation. SIAM Rev 31(3):446–477CrossRefGoogle Scholar
 21.Peletier LA, Benson N, van der Graaf PH (2009) Impact of plasmaprotein binding on receptor occupancy: an analytical description. J Theor Biol 256(2):253–262PubMedCrossRefGoogle Scholar
 22.Krzyzanski W, Wyska E (2008) Pharmacokinetics and pharmacodynamics of erythropoietin receptor in healthy volonteers. Naunyn Schmiedebergs Arch Pharmacol 377:637–645PubMedCrossRefGoogle Scholar
 23.Mager DE, Jusko WJ (2002) Receptormediated pharmacokinetic/pharmacodynamic model of interferonβ 1a in humans. Pharm Res 19(10):1537–1543PubMedCrossRefGoogle Scholar
 24.Aston PJ, Derks G, Raji A, Agoram BM, van der Graaf PH (2011) Mathematical analysis of the pharmacokineticpharmacodynamic (PKPD) behaviour of monoclonal antibodies: predicting in vivo potency. J Theor Biol 281(1):113–121PubMedCrossRefGoogle Scholar
 25.Wagner JG (1971) Biopharmaceutics and relevant pharmacokinetics. Drug Intelligence Publications, Hamilton Press, HamiltonGoogle Scholar