nature publishing group
`
`Eli Lill & c
`EX2268
`y
`o. v. Teva Pharms. lnt'l GMBH
`IPR2018-01422, ·01423, 01424 -01425
`°
`·01426, •01427
`
`I
`
`Monoclonal Antibody Pharmacokinetics and
`Pharmacodynamics
`W Wang1, EQ Wang2 and JP Balthasar3
`
`More than 20 monoclonal antibodies have been approved as therapeutic drugs by the US Food and Drug Administration,
`and it is quite likely that the number of approved antibodies will double in the next 7-10 years. Antibody drugs show
`several desirable characteristics, including good solubility and stability, long persistence in the body, high selectivity
`and specificity, and low risk for bioconversion to toxic metabolites. However, many antibody drugs demonstrate
`attributes that complicate drug development, including very poor oral bioavailability, incomplete absorption following
`intramuscular or subcutaneous administration, nonlinear distribution, and nonlinear elimination.ln addition, antibody
`administration often leads to an endogenous antibody response, which may alter the pharmacokinetics and efficacy of
`the therapeutic antibody. Antibodies have been developed for a wide range of disease conditions, with effects produced
`through a complex array of mechanisms. This article attempts to provide a brief overview of the main determinants of
`antibody pharmacokinetics and pharmacodynamics.
`
`INTRODUCTION
`Antibodies, which are also called immunoglobulins (Igs), are
`large proteins used by the immune system to identify and neu(cid:173)
`tralize foreign objects such as bacteria and viruses. All Ig mol(cid:173)
`ecules are composed of a basic unit of two identical heavy chains
`and two identical light chains, held together by a number of
`disulfide bonds. In humans, there are two types oflight chains
`(KandA.) and five types oflg heavy chains (a., 0, E, y, and fl). 1 Igs
`are grouped into five classes according to the structure of their
`heavy chains: IgA, IgD, IgE, IgG, and IgM. Among these, IgG
`is the predominant class, comprising -80% of the Igs in human
`serum. All of the approved therapeutic antibodies are IgGs, and
`this review focuses on this class.
`Intact IgGs have a molecular weight of -150 kDa and a valence
`of 2 (meaning that each molecule of IgG contains two identical
`antigen-binding domains). The antigen-binding sites are located
`in the complementarity determining regions (CDRs) within the
`Fab portion of the antibody (Figure 1). Fab, which refers to the
`fragment of antigen binding, is composed of domains associated
`with the light chain (VL, CL) and domains associated with the
`heavy chain (VH, CHI). The stem, or Fe, portion oflgG con(cid:173)
`tains the CH2 and CH3 domains of the heavy chains, and this
`region of the antibody is involved with binding to a wide range
`of cell-associated receptors (i.e., Fe receptors). The IgG family of
`
`antibodies may be further divided, again based on the structure
`oftheir heavy chains, into four subclasses: IgG 1, IgG2, IgG3, and
`IgG4. Structural differences among IgG heavy chains lead to dif(cid:173)
`ferences in subclass binding to Fe receptors and, consequently,
`to subclass-specific differences in processes mediated by Fe
`receptors (e.g., activation of complement or antibody-dependent
`cell-mediated cytotoxicity). For example, antibody-dependent
`cell-mediated cytotoxicity by mononuclear cells is more efficient
`for IgG 1 and IgG3 than for IgG2 and IgG4. On the other hand,
`IgG4 is much more active in recruiting the alternative comple(cid:173)
`ment pathway than are the other three IgG subclasses. 1
`Antibody drugs typically possess several desirable pharma(cid:173)
`cological characteristics, such as long serum half-lives, high
`potency, and limited off-target toxicity. Initial antibody thera(cid:173)
`pies were prepared from hyperimmune sera, collected follow(cid:173)
`ing immunization of animals. The resulting antibody product,
`which is derived from a large number of genetically distinct
`cells, contains a distribution oflg isotypes and affinities. In 1975,
`Kohler and Milstein demonstrated that antibody-producing
`B lymphocytes may be fused with myeloma cells to generate
`hybrid cells (hybridomas) that propagate indefinitely in culture
`and secrete antibody. 2 Cloning the hybridoma cells enabled effi(cid:173)
`cient production of antibody derived from a single progenitor,
`and the resulting monoclonal antibody (mAb) preparations are
`
`1 Department of Drug Metabolism and Pharmacokinetics, Merck Research Laboratories, West Point, Pennsylvania, USA; 2Department of Pharmacokinetics, Dynamics,
`and Drug Metabolism, Pfizer Global Research and Development, Groton Laboratories, Groton, Connecticut, USA; 3Department of Pharmaceutical Sciences, Center for
`Protein Therapeutics, School of Pharmacy and Pharmaceutical Sciences, University at Buffalo, The State University of New York, Buffalo, New York, USA.
`Correspondence: JP Balthasar (jb@buffalo.edu)
`
`Received 17 July 2008; accepted 30 July 2008; advance online publication 10 September 2008. doi: 1 0 1 038/clpt.2008.170
`
`548
`
`VOLUME 84 NUMBER S i NOVEMBER 20081 www.nature.com/cpt
`
`

`

`ANTIBODY ABSORPTION
`The majority of marketed antibodies are labeled for intravenous
`(IV) administration; however, several antibodies have been
`approved for extravascular administration. For example, cer(cid:173)
`tolizumab pegol, adalimumab, efalizumab, and omalizumab are
`all approved for subcutaneous (SC) administration. Palivizumab
`is approved for intramuscular (IM) administration, and ranibi(cid:173)
`zumab is administered by intravitreal injection. Antibodies have
`not been successfully developed for oral administration, as oral
`absorption of antibody is limited by presystemic degradation in
`the gastrointestinal tract and by inefficient diffusion or convec(cid:173)
`tion through the gastrointestinal epithelium. With the exception
`of ranibizumab, where intravitreal administration is employed to
`promote a regional effect, efficacy of mAbs following extravascu(cid:173)
`lar administration is dependent on systemic absorption.
`Primary pathways for systemic absorption include convec(cid:173)
`tive transport of antibody through lymphatic vessels and into
`the blood, and diffusion of antibody across blood vessels dis(cid:173)
`tributed near the site of injection. Based on work conducted
`by Supersaxo et at.S that investigated the lymphatic uptake of a
`variety of proteins following SC injection in sheep, it has been
`suggested that the majority of antibody administered via SC
`or IM injection is absorbed via convection through lymphatic
`vessels. However, recent investigations conducted in rats suggest
`that the role of diffusion into blood vessels may be underesti(cid:173)
`mated by the sheep studies.6 Using insulin, bovine serum albu(cid:173)
`min, and erythropoietin as model proteins, Kagan et al. found
`that <3% of the administered dose of each protein was absorbed
`via the lymph. Neither Kagan et al. nor Supersaxo et al. have
`thoroughly investigated the fate oflgG following SC injection
`and, consequently, there is substantial uncertainty regarding the
`primary determinants of antibody absorption. The kinetics of
`antibody absorption, however, has been well described. After
`IM or SC injection, absorption proceeds slowly, and the time
`to reach maximal plasma concentrations (tmax) typically ranges
`from 2 to 8 days. Absolute bioavailability is generally reported
`between 50 and 100%.3
`In practical terms, bioavailability is determined by the rela(cid:173)
`tive rates of presystemic catabolism and systemic absorption.
`Presystemic catabolism may be dependent on rates of extra(cid:173)
`cellular degradation (e.g., via proteolysis), rates of antibody
`endocytosis (e.g., receptor-mediated, fluid phase), and rates
`of recycling through interaction with the Brambell receptor
`(FeRn). FeRn protects IgG from intracellular catabolism, and
`FeRn has been shown to be capable of transporting IgG across
`cell monolayers in both the apical-to-basolateral and basolateral(cid:173)
`to-apical directions. Work from the Balthasar Laboratory
`(A. Garg, P.J. Lowe, and J.P. Balthasar, unpublished data) has
`indicated that the systemic bioavailability of7E3, a monoclonal
`IgG 1 antibody, was threefold higher in wild-type mice vs. FeRn(cid:173)
`deficient mice (82.5 ± 15.6% vs. 28.3 ± 6.9%, P < 0.0001). It
`is not yet known whether the effects of FeRn on SC bioavail(cid:173)
`ability are primarily related to FeRn-mediated protection from
`catabolism or from FeRn-mediated transport across the vascu(cid:173)
`lar endothelium (from interstitial fluid to the blood); however,
`the former mechanism is considered to be more plausible.
`
`In some cases, an inverse relationship between SC bioavail(cid:173)
`ability and antibody dose has been noted? Such relationships
`are suggestive of saturable endocytosis and/or saturable deg(cid:173)
`radation processes. Degradation at the injection site is likely to
`account for some presystemic loss of antibody, but the quantita(cid:173)
`tive significance is uncertain. Charman et al. have demonstrated
`that the major determinant of the SC bioavailability of human
`growth hormone in sheep is presystemic catabolism during the
`course oflymphatic transport. 8 The role oflymphatic catabo(cid:173)
`lism on the bioavailability of other proteins, including mAbs,
`is not known.
`As a result of limited solubility of antibodies in solution
`( -100 mg/ml) and limitations on the volume of fluid that may
`be tolerated with IM or SC injection ( -5 and 2.5 ml, respec(cid:173)
`tively), IM and SC administration are feasible only for anti(cid:173)
`bodies that demonstrate relatively high dose potency. Use of
`multiple injections may help to overcome this limitation, at least
`to some extent. For example, doses of 375 mg of omalizumab
`are routinely administered clinically, via three separate 1-ml SC
`injections.
`Although they have not yet been employed in routine clinical
`use, there is substantial interest in the development of antibodies
`and Fc-fusion proteins for pulmonary delivery.9 The lungs have
`a very large surface area and high perfusion rate. In addition,
`pulmonary epithelial cells are known to express FeRn, which
`may facilitate efficient systemic absorption of antibody delivered
`to the lung. As discussed with SC and IM administration, the
`feasibility of pulmonary delivery of antibodies is likely limited to
`those antibodies associated with very high dose potency, as only
`small volumes of fluid may be delivered to the lung.
`
`ANTIBODY DISTRIBUTION
`The distribution of mAbs is determined by the rate of extra vasa(cid:173)
`tion in tissue, the rate of distribution within tissue, the rate and
`extent of antibody binding in tissue, and the rates of elimination
`from tissue. For large, polar substances such as mAbs, diffusion
`across vascular endothelial cells is very slow, and convection
`is believed to be the primary mechanism responsible for the
`transport of antibody from blood fluid to interstitial fluids of
`tissue. Of note, physiologically based analyses of antibody dis(cid:173)
`position in mice suggest that >98% of antibody enters tissue via
`convection. 10 The rate of extravasation by convective transport,
`or the movement of antibody into tissue by "solvent drag;' is
`determined by the rates of fluid movement from blood to tissue
`and by the sieving effect of paracellular pores in the vascular
`endothelium. Sieving is thought to be largely determined by
`the size and tortuosity of the pores and by the size, shape, and
`charge of the solute (i.e., the antibody). Most physiologically
`based models of antibody disposition describe the uptake clear(cid:173)
`ance for antibody extravasation as a product of the lymph flow
`rate (L) and an efficiency term ( 1 - cr). The reflection coefficient,
`0", represents the fraction of solute sieved during the movement
`of solvent through a pore. In the case of mAbs, tissue reflection
`coefficients are often assumed to be equal in all tissues, with
`values in the range of 0.95-0.98. 10- 12 However, it is likely that
`reflection coefficients may be much lower in tissues such as the
`
`550
`
`VOLUME 84 NUMBER Si NOVEMBER 20081 www.nature.com/cpt
`
`

`

`substances with several repeated epitopes, may bind with two
`or more antibodies, leading to the formation oflarge complexes
`that may be rapidly eliminated by phagocytosis. Elimination of
`large immune complexes may explain, in part, the nonlinear
`elimination kinetics of omalizumab and denosumab, which are
`thought to interact with soluble targets (IgE and receptor acti(cid:173)
`vator of nuclear factor-KB ligand). The majority of marketed
`antibodies demonstrate dose-dependent elimination consistent
`with target-mediated elimination, where clearance decreases as
`a function of dose (e.g., trastuzumab, rituximab, gemtuzumab,
`and panitumumab). Examples of state-of-the-art mathemati(cid:173)
`cal modeling of target-mediated antibody elimination include
`reports by Ng et al., describing the nonlinear disposition of
`TRX1, an anti-CD4mAb; 16 by Hayashi et al., describing the
`nonlinear disposition of omalizumab, an anti-IgE mAbY and
`by Lammerts van Bueren et al., presenting an interesting con(cid:173)
`ceptual model of target-mediated antibody elimination from a
`peripheral distribution compartment. 18
`IgG antibodies may also interact with Fe-y-receptors (FcyR),
`and IgG-FcyR complexes may trigger endocytosis and catabo(cid:173)
`lism. Considering the relatively high affinity ofigG for FcyR
`and the high endogenous concentrations of IgG in plasma
`( -65 f!mol/1), it has been argued that FcyR-mediated elimination
`is unlikely to be important for monomeric IgG. 3 It is possible
`that FcyR-mediated elimination is significant, and perhaps dom(cid:173)
`inant, in cases where antibody is able to form soluble immune
`complexes containing three or more IgG molecules, as well as
`in cases where antibody binds to cells suspended in blood or
`other body fluids (perhaps including viruses, bacteria, plate(cid:173)
`lets, erythrocytes, and leukocytes). IgG "opsonized" particles are
`rapidly engulfed following engagement ofFcyR on macrophages
`and on other phagocytic cells. This mechanism of elimination
`is well supported by the immunology literature; however, little
`work has been performed to link FcyR-mediated phagocytosis
`to the systemic pharmacokinetics of therapeutic antibodies.
`Additional study is required to allow meaningful discussion of
`the role ofFcyR-mediated endocytosis in the elimination of such
`antibodies.
`IgG, like other proteins found in plasma and interstitial
`fluid, may enter cells in all tissues via fluid-phase endocytosis.
`Interestingly, however, IgG differs from most proteins in that
`a significant fraction of endocytosed IgG is not sorted to the
`lysosome but is redirected to the cell surface and released into
`plasma or interstitial fluids. The recycling of IgG is mediated
`by the Brambell receptor, FeRn, which binds to IgG with pH(cid:173)
`dependent affinity. 19•20 Within the acidified environment of the
`early endosome, IgG binds tightly to FeRn. The IgG-FcRn com(cid:173)
`plexes are not delivered to the lysosome for catabolism but rather
`are sorted to the cell surface for fusion with the cell membrane.
`The receptor shows virtually no affinity for IgG at physiological
`pH and, upon fusion of the sorting vesicle with the cell mem(cid:173)
`brane, IgG dissociates from the receptor and is rapidly released
`into extracellular fluid.
`FeRn-mediated recycling of IgG appears to be quite effi(cid:173)
`cient based on studies conducted with knockout mice. In ani(cid:173)
`mals lacking expression of FeRn, IgG clearance is increased
`
`by approximately tenfold,20 which would be consistent with a
`recycling efficiency of90% (i.e., in wild-type animals express(cid:173)
`ing FeRn). Because FeRn expression is limited, FeRn-mediated
`recycling is capacity limited. The average concentration of IgG
`in plasma in humans is -10 mg/ml. At this concentration, IgG
`has a half-life of-25 days21 and a plasma clearance of -10 ml!h
`( -3.5ml!kg/day). High concentrations ofigG are able to saturate
`the recycling system, decreasing recycling efficiency and leading
`to an increase in the fractional catabolic rate ofigG. For exam(cid:173)
`ple, in myeloma patients, where IgG concentrations in plasma
`may approach 100mg/ml, IgG half-life decreases to 8-10 days.
`Conversely, in patients with very low plasma concentrations of
`IgG, the half-life ofigG antibody may be> 70 days.21
`IgG affinity for FeRn is species specific. Human FeRn shows
`high affinity for human IgG and also for IgG from guinea pigs
`and rabbits; however, the human receptor shows very little affin(cid:173)
`ity for IgG derived from most other species, including mice and
`rats.22 The low affinity of human FeRn for mouse IgG helps to
`explain the very rapid elimination of murine mAbs in humans.
`Approved murine monoclonal IgGs (e.g., muromononab-CD3,
`ibritumomab) demonstrate half-lives of -1 day in patients,
`whereas human IgG is typically associated with a half-life of
`-25 days.
`Although FeRn recycling is capacity limited, significant altera(cid:173)
`tion in the efficiency of FeRn recycling is not typically achieved
`with therapeutic doses of mAbs. Most mAbs are administered at
`doses of< 10 mg/kg, which will increase the total IgG "body load"
`by< 1-2%, as humans typically possess 50-100 g of endogenous
`IgG. However, high-dose intravenous immunoglobulin (IVIG)
`therapy, which calls for the administration of 2 g/kg of pooled
`human IgG, increases IgG plasma concentrations sufficiently to
`increase IgG clearance approximately threefold. 23 This increase
`in IgG clearance leads to a decrease in endogenous antibody
`concentrations; consequently, IVIG therapy for the treatment of
`autoimmune conditions may achieve effects by decreasing the
`plasma concentrations of endogenous, pathogenic autoantibod(cid:173)
`ies. Although IVIG therapy is an effective treatment of a variety
`of autoimmune conditions, it is very expensive because of the
`high doses of antibody required. Of note, preclinical experiments
`have demonstrated that anti-FeRn antibodies are able to achieve
`effects similar to those ofiVIG therapy, at dose levels that are
`-100-fold lower than those required for use in IVIG therapy.24
`There is significant interest in the development of specific FeRn
`inhibitors for use in the treatment of autoimmunity.25
`
`IMMUNOGENICITY
`Any exogenous protein may be viewed by the body as foreign
`and trigger immune responses that lead to the generation of
`endogenous antibodies against the protein. Therapeutic anti(cid:173)
`bodies are no exception. mAb drugs may be categorized as
`(i) rodent antibodies, which are typically obtained from murine
`or rat hydridomas; (ii) chimeric antibodies, which are derived
`from chimeras that have been engineered to express IgG antibod(cid:173)
`ies with human constant regions and rodent variable regions; (iii)
`CDR-grafted antibodies, which contain specific regions within
`rodent variable domains, the CDRs, grafted onto a human IgG
`
`552
`
`VOLUME 84 NUMBER Sl NOVEMBER 20081 www.nature.com/cpt
`
`

`

`function was employed to describe the formation of antibody(cid:173)
`ligand complexes, and complexes dissociated via a first-order
`process. The models are nonlinear with respect to antibody(cid:173)
`ligand binding because of the second -order nature of the binding
`process, as well as capacity limitations associated with the avail(cid:173)
`able concentrations ofligand and antibody. The models relate
`unbound ligand concentration to the effect of interest; as such,
`the models link the PK effects of the anti -ligand antibody to the
`pharmacodynamics of the ligand. Of note, the models differ sub(cid:173)
`stantially in terms of their characterization of the fate of the anti(cid:173)
`body-ligand complex. In the infliximab model, it was assumed
`that the antibody-ligand complex was eliminated with the same
`fractional catabolic rate as the unbound ligand, tumor necrosis
`factor-a. Fitting the model parameters to the data resulted in an
`estimation of a 30-40-day half-life for tumor necrosis factor-a,
`which is considerably different from the known value ( < 1 h).
`The omalizumab model, which is much more plausible, does not
`assume an equivalent elimination rate constant for the complex
`and the ligand (IgE) but allows for kinetically distinct elimination
`ofigE, omalizumab, and the IgE-omalizumab complex.
`The recent work ofMarathe et al., which describes the PK/
`PD of denosumab, a monoclonal IgG2 antibody directed
`against the receptor activator of nuclear factor-KB ligand, rep(cid:173)
`resents the state of the art in modeling immunotoxicothera(cid:173)
`pies (Figure 2). 40 The receptor activator of nuclear factor-KB
`ligand is thought to be a soluble ligand, but there is some pos(cid:173)
`sibility that the protein is also expressed on the cell surfaces.
`Denosumab pharmacokinetics were captured with a target(cid:173)
`mediated disposition model, and denosumab pharmacody(cid:173)
`namics were described with a model that relates the unbound
`concentrations of denosumab to the inhibitory effect of the
`antibody on the receptor activator of nuclear factor-KB ligand
`binding. This mechanistic model provided an excellent descrip(cid:173)
`tion of the pharmacokinetics and pharmacodynamics of deno(cid:173)
`sumab in multiple myeloma patients.
`Several antibodies, including rituximab, cetuximab, and tras(cid:173)
`tuzumab, are designed to bind to cell-surface proteins to mediate
`
`Denosumat>-RANKL
`
`II
`~
`
`Active _L__ RANKL
`losreoblaSIS)
`
`OPG-RANKL
`
`rc;;,;oclas~
`r~u~
`
`l
`
`RANK-RANKL
`
`"A;;;
`(osteoclasiS 1
`Serum NXT ~~~~- -
`..
`
`such as volume of each compartment (organ) and blood flow
`rate, are used to build the model and predict human pharma(cid:173)
`cokinetics. A significant advantage of the PBPK approach is that
`it allows prediction of antibody levels in many tissues, including
`tumor. PBPK models are ideally suited to the consideration of
`effects of saturable processes (e.g., target binding, FeRn process(cid:173)
`ing) on antibody pharmacokinetics, and these models are also
`well suited to predict the influence of a variety of factors (e.g.,
`antigen expression, antibody affinity) on the tissue selectivity of
`antibody disposition. Recent PBPK models have incorporated
`FeRn -antibody binding, allowing consideration of the effects
`of FeRn on antibody catabolism and distribution. 11 •12 The limi(cid:173)
`tations for use of PBPK models to predict the disposition of
`antibodies in humans are significant, however. PBPK models
`are complex, mathematically difficult to construct, poorly suited
`to population analyses, and often limited because of a lack of
`tissue concentration data, parameter availability, or parameter
`identifiability.
`Despite the large number of antibodies in development, only
`a handful of reports using preclinical data to predict the clinical
`pharmacokinetics of antibodies have been published, perhaps
`indicating the difficulties associated with the interspecies scaling
`of antibody disposition. Any effort to predict human pharmacoki(cid:173)
`netics based on preclinical disposition data should consider pos(cid:173)
`sible species differences in the expression or turnover of the target
`receptor, antibody affinity for the target, antibody-FeRn binding,
`endogenous IgG concentrations (i.e., as a determinant of FeRn
`saturation), and potential effects of host anti-drug antibodies.
`
`PHARMACODYNAMICS
`mAbs have been marketed for use in the treatment of a wide
`range of conditions, including cancer, autoimmunity, and
`inflammatory diseases. It is convenient to discuss antibody
`pharmacodynamics relating to four main categories of applica(cid:173)
`tions: (i) immunotoxicotherapy, where antibody is employed to
`alter the pharmacokinetics and pharmacodynamics of soluble
`ligands (e.g., drugs, xenobiotics, and cytokines); (ii) elimination
`of target cells; (iii) alteration of cellular function (e.g., receptor
`blockade); and (iv) targeted drug delivery.3
`Antibodies used for immunotoxicotherapy include bevacizu(cid:173)
`mab, adalimumab, ranibizumab, omalizumab, and infliximab.
`Each of these antibodies binds to a soluble ligand (e.g., vascular
`endothelial growth factor or tumor necrosis factor) and alters
`the pharmacokinetics and pharmacodynamics of the ligand.
`These "neutralizing" antibodies act as competitive inhibitors of
`ligand-receptor binding, shifting ligand concentration-effect
`relationships. In addition, by binding to soluble ligand, immu(cid:173)
`notoxicotherapies often produce dramatic alterations in ligand
`pharmacokinetics. In most cases, the anti-ligand antibody will
`decrease the unbound fraction of ligand in plasma, decrease the
`ligand volume of distribution and clearance, and increase the half(cid:173)
`life of the ligand. For example, omalizumab, an anti-IgE mAb,
`dramatically decreases the clearance of its target ligand, leading
`to a fivefold increase in the plasma half-life ofigE.
`PK/PD models for omalizumab and infliximab have been pub(cid:173)
`lished recently. 38•39 In each model, a second-order association
`
`Figure 2 Pharmacodynamic model for denosumab. Marathe eta/. provide
`an excellent example of a pharmacodynamic model for an antibody acting
`as an antagonist of a soluble ligand.40 Denosumab, like other antibodies
`used for immunotoxicotherapy, binds to a soluble ligand (receptor activator
`of nuclear factor-KB ligand, RANKL), preventing the ligand from binding to
`its endogenous receptor (receptor activator of nuclear factor-KB, RANK),
`and antagonizing the effect of the ligand (i.e., inhibiting RANKL stimulation
`of osteoclast maturation). The Marathe eta/. model, which has been
`simplified herein, employs equilibrium binding functions to relate plasma
`concentrations of denosumab, RANK, and the natural RANKL antagonist
`(osteoprotegrin, OPG) to unbound concentrations of RANKL, and to the
`measured biomarker (serum N-telopeptide, NTX).
`
`554
`
`VOLUME 84 NUMBER Si NOVEMBER 20081 www.nature.com/cpt
`
`

`

`V m • efalizumab.CD11 a
`Vc • Km + elalizumab
`
`CD11a
`production
`
`Figure 4 Pharmacodynamic model forefalizumab. The Ng eta/. model,
`selected as an example model for antibodies that alter cellular function,
`describes the effects of efalizumab on CD11 a expression, elimination, and
`the psoriasis area and severity index (PAS I). The model allowed excellent
`characterization of efalizumab pharmacodynamics in a large, population
`pharmacokinetic and pharmacodynamic analysis. The figure shown is a
`simplified version of the model presented by Ng eta/.48
`
`the psoriasis area and severity index, an efficacy end point for
`psoriasis (Figure 4).48
`Preclinical modeling examples include a report by Luo et al.
`that describes the development of a model of cetuximab PK/
`PD using data collected from studies conducted with a murine
`colon carcinoma xenograft model. 49 Although the antibody
`demonstrates nonlinear, target-mediated disposition in humans,
`cetuximab pharmacokinetics were dose-proportional in the
`mouse model and were well characterized with a linear, one(cid:173)
`compartment model. Cetuximab's effects on the phosphoryla(cid:173)
`tion of the epidermal growth factor receptor were captured with
`an indirect effect model, which allowed comparison between
`estimated values of EC50 and EC90 (half maximal effective
`concentration and 90% effective concentration, respectively),
`plasma concentrations of cetuximab achieved in patients, and
`the efficacy of cetuximab in clinical trials.
`In comparison with the other main categories of antibody
`usage, relatively little success has come from the development
`of antibodies for targeted drug delivery. Most of the interest
`in this area has centered on the development of conjugates
`of antibodies and toxic agents (e.g., chemotherapeutic drugs,
`radioisotopes, and biological toxins), with the intent of using
`the high specificity and selectivity of antibodies to mediate
`targeted delivery of toxins. The antibody-toxin conjugates, or
`immunotoxins, carry the complexities shared by other types of
`antibody drugs (e.g., potential for nonlinear target-mediated
`disposition, immunogenicity) . In addition, off-target toxic(cid:173)
`ity is often a greater concern for immunotoxins because of
`the potential for dissociation, in vivo, of the toxin from the
`antibody and because of the high potency of toxins employed.
`Considerable toxicity often results from "nonspecific" distribu(cid:173)
`tion of the immunotoxin to off-target sites. Bone marrow stem
`cells are particularly susceptible to toxicity from immunoto(cid:173)
`xins because of their rapid growth rate and high sensitivity to
`chemotherapy, along with the leaky vasculature of the bone
`marrow, which allows relatively efficient convective uptake of
`immunotoxins.
`
`Most of the work associated with the use of antibodies for
`targeted drug delivery has been focused on the treatment of
`solid tumors. Solid tumors are problematic targets for anti(cid:173)
`body drugs, as tumor growth often leads to the collapse oflym(cid:173)
`phatic vessels within the tumor, which leads to an increase in
`the tumor interstitial pressure. High interstitial pressure mini(cid:173)
`mizes the blood-to-tumor hydrostatic pressure gradient, and
`this decreases the driving force for antibody uptake into tumor
`by convection. Once antibody extravasates, distribution may be
`limited by the binding-site barrier (discussed above), further
`reducing the effectiveness of antibody-directed delivery of toxins
`to solid tumors.
`For chronic immunotoxin therapy, it may be important to
`select a cellular target that is easily accessed by antibody in blood
`(i.e., hematological cells, cells in tissues with "leaky" vascula(cid:173)
`ture), antibodies with little risk for immunogenicity, toxins with
`little risk for immunogenicity (e.g., protein toxins such as ricin
`would not be desired), and conjugation chemistry that allows for
`little off-target release of toxin, but where there is efficient release
`of toxin within target cells (i.e., in cases where this is required for
`efficacy). Successfully marketed antibodies include gemtuzumab
`ozogamicin, tositumomab, and ibritumomab tiuxetan. In each
`case, the antibodies target hematological cells. Tositumomab
`and ibritumomab utilize radioisotope toxins, where dissociation
`from the antibody is not required for the desired cytotoxic effect.
`Gemtuzumab ozogamicin employs a calicheamicin derivative
`toxin that is released in target cells after binding of gemtuzu(cid:173)
`mab to the target receptor (CD33) and after receptor-mediated
`endocytosis of the immunotoxin. The toxin migrates to the
`nucleus and binds DNA, leading to double-strand breaks and
`cell death.
`There are few publications of PK/PD models for immuno(cid:173)
`toxin therapies. Ideally, mathematical models of immunotoxin
`pharmacokinetics and pharmacodynamics should account for
`the intact immunotoxin, "naked" antibody (i.e., antibody alone,
`following release of the toxin), and "free" toxin. In an interest(cid:173)
`ing example, Zhu et al. applied physiologically based modeling
`and simulation to investigate relationships between the dose of
`radioimmunotoxins and uptake of the conjugates into tissue. 50
`Their modeling led to the conclusion that Fab fragments would
`be preferred for use in detection of tumors, whereas Fab2 frag(cid:173)
`ments were predicted to be more effective for use in radioimmu(cid:173)
`notherapy. 50 The structure of their PBPK model may be easily
`adapted to the prediction and characterization of the PK/PD of
`additional immunotoxins.
`
`CONCLUSIONS
`Antibody drugs demonstrate unique, complex PK characteristics.
`Absorption following IM or SC administration is slow and, for
`some antibodies, dose dependent. Antibody distribution kinetics
`is influenced by rates of convective transport, binding to tissue
`sites, and rates of catabolism within tissue. Traditional noncom(cid:173)
`partmental analyses and mammillary models may underestimate
`the steady-state distribution volume of many antibodies, partic(cid:173)
`ularly those associated with substantial elimination from tissue
`sites .. Antibodies often demonstrate target-meditated disposition,
`
`556
`
`VOLUME 84 NUMBER 5 I NOVEMBER 20081 www.nature.com/cpt
`
`

`

`42. Meijer, R.T., Koopmans, R.P., ten Berge, IJ. & Schellekens, P.T.
`Pharmacokinetics of murine anti-human CD3 antibodies in man are
`determined by the disappearance of target antigen.). Pharmacal. Exp. Ther.
`300, 346-353 (2002).
`43. Mould, D.R. eta/. A p