`0312-8963/08/0007-0417/$48 00/0
`REVIEW ARTICLE
`
`© 2008 Adis Data information BV Ali nghts reserved
`
`Pharmacokinetic/Pharmacodynamic Modelling
`in Diabetes Mellitus
`
`Cornelia B. Landersdorfer and William J. Jusko
`
`Department of Pharmaceutical Sciences, School of Pharmacy and Pharmaceutical Sciences, University at Buffalo,
`State University of New York, Buffalo, New York, USA
`
`Contents
`
`ADSHOCT.EE REE EE CELE REEL EE EEE LEE EE EE EEE REECE EOE DED EERE EEE ED EEE REE EES 417
`1. The GluCOSE-INSUIIN SYSTEM0. REE ERE E REEL EEE EERE ER ERED EDR DDE DEED DE EERE DEES EEE EES 418
`2. Mechanisms of Action of Antidiabetic Drugs... ene ER eee Cede E RR eee eee EEE RES 42)
`3. Overview of Diabetes Mellitus Models... nn eee EEE EEE ENED DEE EE De EEE Ede E eb eee EES 422
`4, Models for the Intrinsic Interaction Between Glucose dnd INSUIIN 6... ene ee nen enter eee nnnnes 423
`
`4.1 Models for Diagnostic Tests0.EE EEE E NEEL ERED EEN EEE E ERE EEE eee EE EEE eee EEE 423
`4.2 Models for the Glucose-insulin System Not Intended for Diagnostic Tests... eer ete eter e teeter nnn 426
`4.3 Models for Important Components of the Glucose-Insulin System . 0... eee ence ee ree eee e beeen ens 429
`5. Models for the Glucose-Insulin System Incorporating Drug Effects 0... ccc terete teen een teen eter enseneegs 429
`5,1 Bioplhase Models ..0. 06. icc eee TERRE EE EEE TEER EE ERED EEE ENED HEED CREED DG EG Hee EEE EERE 430
`6.2 The Minimal Model... 1. ne eee eer REDD DEE EERE HEURES ERED ERD EEG FEN EED SUE E EE TEES EES 431
`5.3 Indirect Response Models 20... . 0 ccc te renee ne RDU eS ERED SERED EDUC E RD CEU EEE D EtG Fd eee ede ab baa 432
`5.4 Use of Biophase Models versus Indirect Response Models ....... 0.0: cece rece cere cern eee eet ene ves aeerneenesureees 436
`6, Models that Describe Secondary Drug Effects 2...EE eet nn eee een 437
`7, Models that Describe Effects on Ancillary Blomarkers . 0... eee t enna eee ner eeen env eregeuanenaneneeea 438
`8. Models for Disease Progression oo.ee EEE RE REE EEE EERE EEN EE EEE EEE E EERE EEE REEDED EES 440
`9, Future Needs GNd Perspectives... ee EEE EEE DERE DETER E ENE EE EEE EEE EEE EEE EEE EEE EES 442
`TO. CONMCIUSIONS 2. ee eee EEE REDD TEETER EE ELEN EEE ERE EE ENED EEE DOE EEE ERD R REED REDE EDT R a 443
`
`Abstract
`
`Diabetes mellitus is a major health risk in many countries, and the incidence rates are increasing. Diverse
`therapeutic agents are applied to treat this condition. Since 1960, numerous mathematical models have been
`developed to describe the glucose-insulin system, analyse data from diagnostic tests and quantify drug effects.
`This review summarizes the present state-of-the-art in diabetes modelling, with a focus on models describing
`drug effects, and identifies major strengths and limitations of the published models.
`For diagnostic purposes, the minimal model has remained the most popular choice for several decades, and
`numerous extensions have been developed. Use of the minimal model is limited for applications other than
`diagnostic tests. More mechanistic models that include glucose-insulin feedback in both directions have been
`applied. The use of biophase distribution models for the description of drug effects is not always appropriate.
`Morerecently, the effects of various antidiabetic agents on glucose and insulin have been modelled with indirect
`response models. Such models provide good curvefits and mechanistic descriptionsof the effects of antidiabetic
`drugs on glucose-insulin homeostasis. These and other types of models were used to describe secondary drug
`effects on glucose andinsulin, and effects on ancillary biomarkers. Modelling of disease progression in diabetes
`can utilize indirect response models as a disturbance of homeostasis.
`Future needs are to include glucose-insulin feedback more often, develop mechanistic models for new drug
`groups, consider dual drug effects on complementary subsystems, and incorporate elements of disease progres-
`sion.
`
`MPI EXHIBIT 1051 PAGE1
`
`MPI EXHIBIT 1051 PAGE 1
`
`
`
`418
`Landersdorfer & Jusko
`
`Diabetes mellitus is an increasing health problem in many
`countries. In 2006, the WHOestimated that 180 million people
`worldwide had diabetes, and that this number will likely be more
`than doubled by 2030. The number of deaths attributable to
`diabetes has been estimated at 2.9 million per year!!! Type 1
`diabetes mellitus (T1DM) is characterized by an inability of the
`body to produce insulin, and has to be treated with exogenous
`insulin. Type 2 diabetes mellitus (T2DM)reflects both insulin
`resistance of liver and peripheral tissues, and deficient insulin
`secretion by pancreatic B-cells.'2! Variousoral antidiabetic drugs,
`as well as insulin, are used for treatment. The factors contributing
`to the pathophysiology and disease progression of diabetes are not
`yet fully understood. Long-term complications such as neuro-
`pathy, nephropathy and retinopathy are debilitating for patients
`and result in high healthcare costs. In the US, diabetes was the
`leading cause of renal failure and new blindness in 2005,'3! and
`total diabetes costs were estimated at $US174billion in 2007.41
`
`For these reasons, early diagnosis and adequate treatment are
`important to decrease the long-term adverse effects of diabetes.
`Mathematical models are needed to better understand the glucose-
`insulin system, to evaluate diagnostic tests, to study and predict
`drug effects, and to quantify disease progression. Mechanism-
`based models are preferable, as they can be used to study the
`mechanism of action and to predict the effects of new dosage
`regimens. If the number of time- and cost-intensive long-term
`clinical trials that need to be performed can be reduced by simula-
`tion studies, the time and costs involved in drug development
`could be considerably decreased, and optimized dosage regimens
`could be prospectively designed andclinically evaluated. Pharma-
`cokinetic/pharmacodynamic modelling can also help to design
`better studies, to identify agents with undesirable properties earlier
`and to decreaseattrition rates at late stages of drug development.!!
`Since the 1960s, numerousinteresting and useful mathematical
`models have been developed for various applications and with
`different characteristics and employing different methods. With
`this large number of models. it is sometimes difficult to decide
`which one should be used for a specific purpose. A few reviews
`are available on certain aspects of diabetes modelling.!*'" How-
`ever, no extensive overview has been published that includes the
`different
`types of pharmacodynamic models and assesses the
`mechanismsof drug effects. This review summarizesthe state-of-
`the-art in diabetes modelling, with a focus on models describing
`drug effects. We soughtto identify major strengths and limitations
`of the published models and further needs and perspectives in
`diabetes modelling.
`Weperformeda literature search of MEDLINE, EMBASE,the
`references of published papers and abstracts from various confer-
`ences. The models were subcategorized as models for (1) diagnos-
`tic tests; (ii) intrinsic interactions of glucose and insulin; (iii) an-
`tidiabetic drug effects: (iv) secondary drug effects; (v) ancillary
`
`© 2008 Adis Data Information BV. All nights reserved
`
`biomarkers; and (vi) disease progression. This review focuses on
`models for system and parameter estimation and includes some
`simulation models.
`
`1. The Glucose-insulin System
`
`Diabetes is a multiorgan disease, which is mainly characterized
`by a disturbance of glucose-insulin homeostasis. A simplified
`diagram of the regulation of the glucose system by insulin and
`other endogenous substances is shown in figure 1. The diagram
`describes the action of relevant biomarkers by stimulation or
`inhibition of production or utilization of other endogenous sub-
`stances. The plasma glucose concentration is increased by hepatic
`glucose production and absorption from the gut after food intake.
`It decreases dueto utilization by the brain and peripheral tissues
`such as muscle and fat. One of the key characteristics of the
`glucose-insulin system is reciprocal feedback between glucose and
`insulin, ie. an increase in glucose concentrations stimulates pro-
`duction of insulin, and insulin in turn stimulates utilization of
`glucose and inhibits production of glucose.
`In general, the human body aims to maintain blood glucose
`concentrations within a narrow range of about 80-115 mg/dL.)
`For example, tissues such as the brain and erythrocytes need a
`constant supply of glucose. Therefore, in the fasting state, glucose
`is produced by breakdown of glycogen (glycogenolysis) in the
`liver. After an extended fasting period, hepatic glycogen stores are
`depleted. Then glucose is produced mainly from gluconeogenesis
`in the liver (and to a smail extent in the kidneys), utilizing mainly
`amino acids from muscle tissue, lactate or glycerol. Both glycoge-
`nolysis and gluconeogenesis are stimulated by glucagon,a peptide
`hormonethat is secreted from pancreatic o-cells at low blood
`glucose concentrations (figure 1). Only a low basal insulin secre-
`tion occurs at glucose concentrations below about 80 mg/dL.!"4] In
`the fasting state, most of the glucose is taken up by insulin-
`independenttissues such as the brain, and <20% is taken up by the
`muscle.!'5] During strenuous exercise and anaerobic conditions,
`glycogen in muscle tissue is broken down to lactate, which is
`released into the blood and taken up by the liver, where it can be
`stored as glycogen or converted into glucose by gluconeogenesis.
`The concept of resynthesis of glucose from lactate in the liver was
`formulated by Cori and is knownas the Cori cycle or the glucose-
`lactate cycle.!!6! The Cori cycle includes the breakdown of glucose
`to lactate in muscle tissue via anaerobic glycolysis, uptake of
`lactate into the liver, conversion of lactate to glucose inthe liver,
`and uptake of glucose into muscle tissue again.
`After a meal is ingested, complex carbohydrates from the food
`are split into oligosaccharides which, in turn, are broken down to
`monosaccharides such as glucose by o-glucosidases in the gut
`wall (figure 1). Glucose is absorbed into the blood, and glucose
`concentrations increase. Due to increasing glucose concentrations,
`secretion of insulin from pancreatic B-cells is stimulated (figure 1).
`
`Clin Pharmacokinet 2008, 47 (7)
`
`MPI EXHIBIT 1051 PAGE 2
`
`MPI EXHIBIT 1051 PAGE 2
`
`
`
`glucose
`
`ine /
`
`Plasma
`
`Fig. 1. Simplified diagram of the regulation of glucose metabolism,!""“"3! describing the action of relevant biomarkers by stimulation (open bars) orinhibition
`(solid bars) of production/utilization of other endogenous substances. Many more biomarkers are involved in glucose regulation but have been omitted
`from this diagram for clarity. AG = a-glucosidases; FFA = free fatty acids; GLP-1 = glucagon-like peptide 1.
`
`Atincreased blood glucose concentrations, more glucoseis taken
`up into B-cells and metabolized to adenosine triphosphate, which
`results in closure of potassium channels, membrane depolariza-
`tion, opening of calcium channels, an increase in intracellular
`calcium concentrations and, eventually, increased insulin secre-
`tion, !!!]
`
`Insulin is secreted in equimolar amounts with C-peptide (con-
`necting peptide) from B-cells. C-peptide is formed due to cleavage
`of insulin from pro-insulin. As a large and variable proportion of
`the secreted insulin is metabolized during the first pass through the
`liver, and C-peptide is not extracted by the liver, peripheral C-
`peptide concentrations are often measured to study insulin secre-
`tion (7)
`
`In the short-term, insulin secretion is stimulated by high con-
`centrations of glucose, free fatty acids (FFA)!"®!9! (figure 1) and
`aminoacids.'7°l However, long-term exposure to elevated concen-
`trations of FFA (lipotoxicity)!'9?"! (figure 1), glucose (glucotoxici-
`ty)?! or amino acids!°! results in B-cell failure and decreased
`insulin secretion. Insulin secretion is also stimulated by incretin
`hormones, such as glucagon-like peptide 1 (GLP-1) and glucose-
`dependent insulinotropic polypeptide (GIP), which are secreted
`
`© 2008 Adis Data Information BV.All rights reserved.
`
`from the gut wall after food intake (see later in this section and in
`figure 1).
`
`As shownin figure 1, insulin inhibits hepatic glucose produc-
`tion and stimulates glucose utilization by peripheral tissues. Insu-
`lin promotes uptake of glucose and storage as glycogenin theliver
`and skeletal muscle. It also enhances glucose disposal by stimula-
`tion of glycolysis in muscle tissue. Liver glycogen is a readily
`available glucose reservoir for the maintenance of blood glucose
`concentrations in the fasting state, whereas muscle glycogen is
`used to store energy for the muscle itself. Glucose uptake by the
`brain is insulin-independent, as the brain cannot store glucose and
`therefore needs a permanentsupply of energy (as shownin figure
`1, where there is no stimulation of brain glucose uptake by
`insulin). At a high carbohydrate intake, insulin also stimulates
`conversion of glucose to fatty acids and their uptake into adipose
`tissue, where they are stored as triglycerides. Insulin also stimu-
`lates uptake of aminoacids and their storage as protein in skeletal
`muscle, and uptake of FFA andtheir storage as triglycerides in
`adipose tissue. In addition, insulin inhibits hepatic gluconeogene-
`sis by inhibition of key enzymes of this pathway,
`including
`phosphoenolpyruvate carboxykinase (PEPCK), which catalyses
`the rate-limiting step of gluconeogenesis andalso inhibits hepatic
`
`Clin Pharmacokinet 2008; 47 (7)
`
`MPI EXHIBIT 1051 PAGE 3
`
`PK/PD Modelling in Diabetes Mellitus
`
`419
`
`Appetite
`
`= w
`
`MPI EXHIBIT 1051 PAGE 3
`
`
`
`420
`Landersdorfer & Jusko
`
`glycogenolysis (figure 1). Insulin also inhibits the release of amino
`acids from skeletal muscle and the release of FFA and tri-
`glycerides from adipose tissue.
`These insulin actions lead to a decrease in blood glucose
`concentrations and promote storage of energy in the tissues (ana-
`bolic action). Glucagon, conversely, stimulates the release of
`glucose, FFA and amino acids from the liver, fat and muscle
`(catabolic action). At high blood glucose concentrations, secretion
`of insulin is stimulated while glucagon secretion is inhibited, and
`vice versa at low blood glucose concentrations.
`TIDMpatients have an absolute deficiency of insulin, whereas
`T2DMis characterized by impaired insulin secretion and resis-
`tance to insulin action in the liver and peripheral tissues. T1IDM
`and T2DM patients who require insulin therapy have elevated
`postprandial glucagon concentrations,'-2! whereas T1DM patients
`also have fasting hyperglycaemia. In addition, it has been pro-
`posed that reduced suppression of glucagon secretion by insulin
`contributes to impaired glucose tolerance.!?*! Insulin deficiency
`and glucagon excess lead to decreased glucose uptake into cells,
`increased breakdownof protein and lipolysis. In severe uncon-
`trolled diabetes,
`the release of large amounts of FFA that are
`converted to ketone bodies in the liver can result in ketoacidosis.
`
`Amylin is co-secreted with insulin from B-cells. It is a hormone
`that has been shownto inhibit postprandial glucagon secretion!?“!
`(figure 1). Amylin also acts by delaying gastric emptying and
`reducing food intake.'*2! Therefore, amylin promotes the decrease
`of high postprandial blood glucose concentrations. Amylin secre-
`tion is lacking in TIDM patients and impaired in T2DM patients
`because of depletion or dysfunction of B-cells,'25! In rodents,
`especially in the fasting state,
`injection of amylin resulted in
`increased concentrationsof lactate due to enhanced muscle glyco-
`gen breakdown. Subsequently, glucose concentrations increased,
`as lactate is a substrate for gluconeogenesis in the liver.?67]
`Increased concentrations of lactate in human muscle, adipose
`tissue and plasma,due to lactate release from muscle and adipose
`tissue, have been shownafter glucose ingestion.78!
`After enteral nutrition, the incretin hormones GLP-| and GIP
`are secreted from cells in the gut wall (figure 1). The secretion is
`stimulated by endocrine and neural signals and by direct stimula-
`tion of the intestinal cells by digested nutrients. Both GLP-1 and
`GIP stimulate glucose-dependent insulin secretion from B-cells
`(figure 1), enhance B-cell proliferation and increase B-cell resis-
`tance to apoptosis. In addition, GLP-1 inhibits glucagon secretion,
`slows downgastric emptying and food ingestion, and decreases
`food intake.*°l Therefore, GLP-1 and GIP both act to decrease
`high blood glucose concentrations after food intake. Both GLP-1
`and GIP are rapidly inactivated by the ubiquitous enzyme dipep-
`tidylpeptidase-4 (DPP-4). The half-life of active GLP-1 in plasma
`is <2 minutes.20] The active GIP half-life has been reported to
`be 7 minutes in healthy subjects and 5 minutes in T2DM pa-
`
`tients.2°3"1 Reduced secretion of GLP-1 and decreased action of
`GIP in T2DM patients compared with healthy subjects has been
`reported.29
`
`FFA play an important role in glucose metabolism. FFA are
`important for B-cell function and stimulate insulin secretion.!'®)
`However, permanently increased FFA concentrations cause ‘li-
`potoxicity’ and decrease insulin secretion from B-cells (figure 1).
`Obesity, especially an increased mass of visceral fat, leads to
`increased concentrations of FFA, as visceral fat exhibits higher
`lipolytic activity than subcutaneousfat. Increased concentrations
`of FFA stimulate hepatic gluconeogenesis!*?! and decrease glucose
`uptakeinto skeletal muscle (figure 1). Permanently increased FFA
`concentrations promote insulin resistance in both skeletal muscle
`andthe liver. As skeletal muscle is responsible for about 75-80%
`of insulin-mediated glucose disposal, decreased glucose uptake by
`the muscle due to peripheral insulin resistance can substantially
`increase plasma glucose concentrations.'7*35]
`In 1963, Randle
`et al.'°¢) proposed the glucose-fatty acid cycle, according to which
`fuel selection from FFA is preferred over glucose. Therefore,
`increased availability of FFA results in decreased glucose uptake
`into muscle cells. More recently, it was reported that FFA also
`induce or enhance insulin resistance due to alterations in insulin
`signalling,?”! which leads to decreased muscle glucosetransport,
`possibly due to diminished glucose transporter protein isoen-
`zyme 4 (GLUT-4)translocation.28! GLUT-4 is the major insulin-
`sensitive glucose transporter in adipose tissue and muscle, and
`facilitates glucose uptake into these tissues.5%! In obese and T2DM
`patients,
`lipolysis is increased, as the adipose tissue is insulin
`resistant, therefore plasma FFA concentrations are higher which,
`in turn, aggravates insulin resistance in the liver and skeletal
`muscle, and causes deterioration of B-cell function.
`
`Recently, it was recognized that adipose tissue is not only a
`storage tissue but has endocrine functions and secretes hormones
`(adipokines) such as adiponectin and leptin (figure 1). Adiponec-
`tin, which is specifically secreted from adiposetissue, stimulates
`glucose uptake and utilization in skeletal muscle and inhibits
`gluconeogenesis in the liver (figure 1). Therefore it supports
`insulin action, increases insulin sensitivity and decreases plasma
`glucose concentrations. Reduced adiponectin concentrations have
`been related to a higher incidence of diabetes, insulin resistance,
`cardiovascular disease and metabolic syndrome.!4041
`
`Leptin, which is produced primarily by adipose tissue, acts as a
`central satiety signal (figure 1). Leptin decreases food intake,
`stimulates energy expenditure and increases insulin sensitivity.
`Leptin- or leptin receptor-deficient animal models exhibit obesity
`and insulin resistance. Obese subjects have higher leptin concen-
`trations because they have a larger fat mass than lean subjects.
`However, obesity appears as a condition ofrelative leptin resis-
`tance, and therefore the effect of leptin is decreased.!4!-47!
`
`© 2008 Adis Data Information BV All nghts reserved.
`
`Clin Pharmacokinet 2008, 47 (7)
`
`MPI EXHIBIT 1051 PAGE 4
`
`MPI EXHIBIT 1051 PAGE 4
`
`
`
`421
`PK/PD Modelling in Diabetes Mellitus
`EGSPSPEETETRAUTEETSESESRPESSeSSSSSTEASESDRESSINetESSSRISTRSTSESEEEEESCTIOTTPO
`
`A simplified diagram, such as in figure 1, can be useful for
`model building. Depending on the observed data and the most
`relevant subsystem (e.g. for the action of a drug), part ofthis
`diagram can be used asthe starting point for a model. To date,
`most mathematical models have focused on glucose, insulin or
`both, and have not included other biomarkers.
`
`2. Mechanismsof Action of Antidiabetic Drugs
`
`TIDM patients cannot produce insulin, and therefore their
`treatment mainly consists of exogenous insulin. Numerousthera-
`peutic optionsare available for T2DM. An overview of the mech-
`anismsandsites of action for the different groups of antidiabetic
`drugs is shownin figure 2.
`Hyperglycaemia in T2DM patients can, among other reasons,
`be due to the following: (i) decreased uptake of glucose into
`skeletal muscle due to peripheral insulin resistance; (ii) increased
`hepatic glucose production (mainly gluconeogenesis) due to hep-
`atic insulin resistance; and (iii) decreased insulin secretion due to
`exhaustion of B-cells, genetic causes, lipotoxicity or glucotoxici-
`ty.°-3] Therefore, drugs that ameliorate one or more of these
`defects can be usedin treating T2DM.It is generally assumedthat
`
`insulin resistance precedes B-cell failure in T2DM.!2:'5] However,
`there is also evidence for the theory that B-cell dysfunction is
`present before insulin resistance.'**! In general, as long as the B-
`cells can increase insulin production enough to compensate for an
`enhanced need due to insulin resistance, the patients do not be-
`come hyperglycaemic.4] However, loss of B-cell function, most
`likely due to accelerated apoptosis,'“! leads to overt diabetes.
`Therefore, it has been suggested that approachesthat delay the loss
`of B-cell function, e.g. by decreasing glucotoxicity and lipotoxici-
`ty, which are reversible, should be considered.!”!
`Sulfonylurea drugs stimulate insulin secretion from B-cells.
`Through binding of the drug to receptors on the B-cells, potassium
`channels are closed, whichresults in stimulation of insulin secre-
`tion by a mechanism similar to stimulation of insulin secretion by
`glucose, as described in section 1. Examples of sulfonylurea drugs
`for which models have been proposed are glimepiride, gliclazide,
`tolbutamide and glibenclamide (glyburide). Meglitinides such as
`repaglinide and nateglinide stimulate insulin secretion by the same
`mechanismas sulfonylureas but have a faster onset and a shorter
`duration of action.!"")
`
`Metformin is one of the biguanides. It is reported to suppress
`gluconeogenesis and glycogenolysis,
`increase insulin-mediated
`
`Site
`
`Mechanism ofaction
`
`Antidiabetic agents
`
`Brain
`
`Intestines
`
`Pancreas
`
`Liver
`
`Muscle
`
`Fat
`
`
`
`|
`
`Delayof gastric emptying
`
`Delayof glucose absorption
`Stimulation of GLP-1 release
`
`Inhibition of glucagon release
`
`Acute stimulation of insulin release
`
`Stimulation of insulin biosynthesis
`
`Inhibition of B-cell apoptosis
`
`Stimulation of B-cell differentiation
`
`Incretin mimetics
`
`Pramlintide
`
`a-Glucosidaseinhibitors
`
`. Sulfonylureas
`
`Meglitinides
`
`Incretin mimetics/DPP-4 inhibitors
`
`Inhibition of glucose production
`
`
`Increasein hepatic insulin sensitivity
`
` — Regulation of food intake
`
`Metformin
`
`Thiazolidinediones
`
`Coe SuppressionofFFArelease
`
`
`Increase in muscle insulin sensitivity
`
`
`
`
`
`Fat redistribution (visceral to SC)
`
`Modulation of adipokine release
`
`Fig. 2. Sites and mechanismsof action of antidiabetic agents (adapted from Stumvollet al.,'"5) with permission from Elsevier). DPP-4 = dipeptidylpep-
`tidase-4; FFA = free fatty acids; GLP-1 = glucagon-like peptide 1; SC = subcutaneous.
`
`© 2008 Adis Data Information BV.All rights reserved.
`
`Clin Pharmacokinet 2008; 47 (7)
`
`MPI EXHIBIT 1051 PAGE 5
`
`MPI EXHIBIT 1051 PAGE 5
`
`
`
`422
`Landersdorfer & Jusko
`
`glucose uptake into muscle and adipose tissue, and stimulate
`anaerobic glucose metabolism in the intestine. Metformin is some-
`times called an insulin sensitizer, as it
`increases hepatic and
`peripheral insulin sensitivity."
`Thiazolidinediones such as rosiglitazone and pioglitazone act
`by stimulation of nuclear peroxisomeproliferator-activated recep-
`tor y (PPAR), which affects the transcription of genes that regu-
`late glucose and lipid metabolism. Stimulation of PPARYincreases
`glucose uptake by the GLUT-4 transporter in muscle and adipose
`tissue by improving translocation of GLUT-4to thecell surface.
`Stimulation of PPARy also decreases concentrations of FFA,
`whichresults in enhanced glucose utilization, decreased gluconeo-
`genesis and increased muscle insulin sensitivity.
`In addition,
`PPARy agonists stimulate secretion of adiponectin and inhibit
`secretion of tumour necrosis factor (TNF)-o from fat cells, affect
`the secretion of other adipokines, and promoteredistribution of fat
`from visceral to subcutaneous fat, which has lowerlipolytic ac-
`tivity. Thiazolidinediones are also often called insulin sensitizers
`and havebeenreportedto partially reverse insulin resistance.'#5! In
`addition, they improvefasting plasma glucose (FPG)levels, main-
`ly by inhibition of gluconeogenesis.“6!
`Acarbose and miglitol are o-glucosidase inhibitors, which slow
`down the breakdown of poly- and oligosaccharides to monosac-
`charides, such as glucose, and delay glucose absorption from the
`intestines. Glucose is absorbed from more distant regions of the
`gut, which results in increased stimulation of GLP-1 produc-
`tion.!!5! @-Glucosidase inhibitors are not absorbed into the system-
`ic circulation.
`
`Pramlintide is an amylin analogue and amylin receptor agonist.
`It retains the biological activity of amylin but,
`in contrast to
`amylin, it does not aggregate into amyloid plaques, which are
`deposited in B-cells. Amyloid plaques may be linked to B-cell
`dysfunction">! and development of T2DM.!5! Pramlintide is used
`as an adjunct to mealtime insulin in TIDM and T2DMpatients.
`Like amylin, pramlintide decreases postprandial secretion of glu-
`cagon,delays gastric emptying andincreasessatiety,'?247! all lead-
`ing to decreased glucose excursions after a meal.
`Incretin mimetics are GLP-1 receptor agonists that are not
`inactivated by DPP-4 and have a longer duration of action than
`GLP-1. Like GLP-1,
`they stimulate glucose-dependent insulin
`release, stimulate insulin biosynthesis and B-cell differentiation.
`inhibit B-cell apoptosis and enhance satiety.!°?) Examples are
`exenatide (synthetic exendin-4), which is approved in the US and
`the EU, and liraglutide, which is in the late stages of clinical
`development.
`DPP-4 inhibitors prolong the action of GLP-1 by inhibiting its
`inactivation by DPP-4. They have similar effects to those of the
`incretin mimetics but do not seem to influence food intake.!?%!
`Sitagliptin is on the market in the US and the EU. Vildagliptin is
`approved in Europe.
`
`Overall, as diabetes is a multiorgan disease, drugs with differ-
`ent sites and mechanisms of action are used. Furthermore, one
`drug can have more than one site or mechanism of action. If a
`model
`includes these different mechanisms for one drug,
`the
`action of a drug can potentially be better described and predicted.
`However, more data and more complex models are usually
`needed.
`
`3. Overview of Diabetes Mellitus Models
`
`Most of the earlier models focused on description of the glu-
`cose-insulin system and evaluation of diagnostic tests (table I). In
`1980, Berman'*8! published an overview of models for insulin
`kinetics. Models published during the 1960s and 70s have been
`divided into two categories:'*9! (i) ‘simple’ models with a mini-
`mum number of parameters, such as the pioneer model published
`by Bolie in 1961;and (ii) comprehensive ‘library’ models,5!54)
`which aim to describe the system as extensively as possible. For
`the latter, not all parameters can be estimated.'*! Two models for
`diagnostic purposes that were originally published in 1979 and are
`still widely used are the minimal model!**| and homeostatic model
`assessment (HOMA).'*5!
`
`Overall, a large number of models have been developed. Table
`II lists various purposes for which models have been published.
`Models for estimation tend to have fewer parameters to retain
`parameter estimability. They are most often used to estimate
`pharmacokinetic/pharmacodynamic parameters after diagnostic
`tests or administration of antidiabetic drugs, and to quantify dis-
`ease progression. Models for simulation are usually large, too
`complex to allow estimation, and aim to describe the physiological
`situation as closely as possible for certain organsortissues or even
`the whole body. Whole-body simulation models may be used to
`perform virtual trials, which are difficult to perform in real human
`subjects for ethical or practical reasons.
`
`Table I. Early models of the glucose-insulin system
`
`Year
`1939
`
`1959
`1961
`1965
`
`Development
`First approach to measureinsulin sensitivity
`in vivo
`
`References
`56
`
`57
`Steele’s equations for glucose (dis)appearance
`50
`Bolie pioneer model for glucose-insulin system
`Modelsimilar to Bolie’s fitted to data from OGTT 58
`
`59,60
`Glucose clamp technique introduced
`1966
`51-53
`1970-5 More comprehensive‘library’ models
`54
`1979
`Original ‘minimal mode?
`55
`1979
`Original HOMA model
`HOMA = homeostatic model assessment; OGTT = oral glucose tolerance
`test.
`
`© 2008 Adis Data information BV.All nghts reserved.
`
`Ctin Pharmacokinet 2008, 47 (7)
`
`MPI EXHIBIT 1051 PAGE 6
`
`MPI EXHIBIT 1051 PAGE 6
`
`
`
`423
`PK/PD Modelling in Diabetes Mellitus
`
`
`Table Il. Array and purposes of developed models
`Estimation
`
`Model! and purpose
`
`Simulation
`
`only
`
`Diagnostics:insulin sensitivity and pancreatic
`B-cell function
`Glucose-insulin system after tolerance tests
`
`Glucose-insulin system in the physiological
`situation
`
`Important subsystems
`B-cell dynamics
`insulin binding to receptors
`oscillations in insulin secretion
`
`glycosylation of haemoglobin
`fuel selection in muscle tissue
`
`insulin signalling
`translocation of GLUT-4
`
`/
`
`v
`
`v
`
`/
`v
`
`v
`
`Effects of antidiabetic drugs
`Secondary drug effects
`Ancillary biomarkers
`Disease progression
`Whole-body simulations (including drug effects)
`GLUT-4 = glucose transporter protein tsoenzyme 4.
`
`S\NON
`
`v
`
`v
`
`NNNNN
`
`N\
`
`v
`
`4. Models for the Intrinsic Interaction Between
`Glucose andInsulin
`
`4.1 Models for Diagnostic Tests
`
`To establish if a person is diabetic or likely to becomediabetic,
`various diagnostic tests have been developed. The goal of these
`experiments is to obtain an estimate of B-cell function, insulin
`sensitivity (most often) or both. Therefore, the clinical utility of
`the models that may be applied to analyse data from the diagnostic
`tests described belowis to indicate the diabetic or prediabetic state
`of a patient. Several methods for estimation of B-cell function
`were compared by Kjemset al.!!7°!] Overviews of methods to
`determine insulin sensitivity were published by Monzillo and
`Hamdy!'*! and Wallace and Matthews.'©! The two methods for
`determination of insulin sensitivity that are considered as ‘gold-
`standard’ by the American Diabetes Association (ADA) are the
`hyperinsulinaemic euglycaemic clamp and the intravenous glu-
`cose tolerance test (IVGTT). For the euglycaemic clamp, general-
`ly a constantrate insulin infusion is given, whichresults in insulin
`concentrations above the physiological baseline. Blood glucoseis
`measured frequently and maintained within the euglycaemic range
`by a glucose infusion with adjusted rates. As hepatic glucose
`production is assumed to be completely inhibited by exogenous
`insulin infusion,the glucose infusion rate (GIR) equals the glucose
`utilization rate and reflects the insulin sensitivity of the peripheral
`tissues. Although some authors have applied or developed models
`
`for glucose clamp data,'®! in the majority of studies, no model-
`ling was performed for evaluation of the experimental data.
`
`For the IVGTT, a glucose bolus of 0.3 g/kg of bodyweightis
`administered, and plasma glucose and insulin are measured at
`12-30 timepoints over 3 hours after the dose.'®! The minimal
`model was developed by Bergman et al.!49-5469.791 to estimate
`insulin sensitivity and glucose effectiveness after an IVGTT. The
`intention wasto find a modelthat could capture the most important
`physiological features of the glucose-insulin system and allow
`estimation of key parameters. The minimal model provides an
`estimate of insulin sensitivity without the need to perform the
`experimentally more complex glucose clamp method. The mini-
`mal model was developed from experiments in five dogs. Seven
`different models were compared using data from nine experiments
`in two dogs r