Review
Foundation
Early integration of pharmacokinetic and dynamic reasoning is essential for optimal development of lead compounds: strategic considerations

https://doi.org/10.1016/j.drudis.2008.12.011Get rights and content

The aims of this report are firstly to raise awareness among kineticists and pharmacologists as to why pharmacokinetic–pharmacodynamic (PKPD) integration is essential for target validation (TV), optimizing development of lead compounds (lead generation [LG] and lead optimization [LO]) and scaling these to human. A related aim is to demonstrate strategic examples of PKPD collaborations that have improved the planning, execution and evaluation of experiments in primary and safety pharmacology. Examples include design of TV studies, design and data ‘pruning’ of PKPD studies in LO, analysis of data with marginal and substantial temporal (time) differences between exposure and response, design of safety pharmacology studies, assessment of safety margin and assessment of uncertainties in predictions of first dose in human.

Introduction

Pharmacokinetics (PK) is the science of the time course of drugs in the organism – that is, it investigates what the body does to the drug. Pharmacodynamics (PD) concerns the study of the time course of biological effects of drugs, the relationship to drug exposure, and drug mechanisms of action – that is, it examines what the drug does to the body. We believe that a better understanding of the inter-relations between PK and PD and sometimes the lack of concordance between the two, might shed light on situations where one or other needs to be optimized in drug discovery and development. It is, therefore, our contention that in vivo pharmacology and in vivo PK must play a greater and more intelligent role in drug discovery if their combined value is to be fully utilized. In recent years, several articles have highlighted the need for drug discovery that is driven by creativity rather than by process or technology (for review see 1, 2). For example, there has been a tendency to focus narrowly on the target and to underestimate the complexity of the physiological role of the target in the intact organism [3]. Binding of a molecule may activate, inactivate or modulate a target's function, and further complications arise if a drug has to simultaneously interact with more than one target to express its therapeutic effect [2]. There are many examples of how process-driven and automated the Drug Metabolism and Pharmacokinetics (DMPK) screen has become, increasing the risk of falling into the trap of ascribing a crucial and central role to a single parameter or just a few parameters (e.g. plasma protein binding, intrinsic hepatic clearance and transcellular flux). Often, for example, industrial scientists state that high plasma protein binding increases the dose size, that the average unbound plasma concentration should be greater than a multiple of three of EC50, or that a short plasma half-life limits the potential for convenient daily dosing in human. Not long ago it was believed by many pharmacokineticists, pharmacologists and physicians that there was no relationship between drug concentration in plasma and the time course of drug action, and that one reason for this was that the pharmacological response often lagged behind the plasma concentration 4, 5, 6. This apparent time lag between response and plasma concentration was first rationalized by [7] and further investigated by [8]. More recently, physiological turnover models have offered better solutions for drugs whose mechanism of action consists of either inhibition or stimulation of a physiological process involved in the clinical expression of drug action 5, 9, 10. Mechanism-based modeling can therefore provide a truely effective link between in vivo pharmacology, DMPK and safety sciences.

The aims of this paper are primarily to explain the rationale behind pharmacokinetic–pharmacodynamic (i.e. quantitative pharmacology, PKPD) reasoning and to raise awareness among pharmacokineticists and pharmacologists of why PKPD integration is essential for target validation (TV), optimizing development of lead compounds (lead generation [LG] and lead optimization [LO]) and scaling these to human. Finally, we provide examples of PKPD collaborations that have improved the design, execution and evaluation of experiments.

Section snippets

What is PKPD reasoning?

Kinetic–dynamic reasoning should, whenever possible, be based on in vitro and in vivo concentration–time, response–time and concentration–response relationships (Fig. 1), with an underlying ambition to couple this to the disease state. In other words, PKPD attempts to describe the biological responses produced by drugs and to define the underlying mechanisms by which the responses are generated. Related to this is the need to address key issues such as temporal differences, exposure–effect

Consequences of ignoring an integrated PKPD approach

What can happen if we fail to use an integrated PKPD approach to experimental design? As we will show, studies may be designed incorrectly, thereby either overestimating or underestimating the risk and/or efficacy of a compound. For example, protein binding across two or more species may differ substantially and even blur information about efficacious concentrations or safety margins. Thus, data should be on the basis of unbound concentrations to avoid significant errors in estimating risk,

PKPD integration during target validation and lead generation

Over the past two decades, the focus on in vitro high capacity assays and a ‘process-led’ way of working has resulted in a less stringent interdisciplinary approach to assessing a compound's combined PKPD characteristics, and more of a ‘tick the box’ approach. As early as TV and LG some compounds prove to be efficacious in vivo, in spite of the fact that their in vitro or in vivo PK properties are not optimal (e.g. high intrinsic clearance, low bioavailability, and/or short half-life). This

PKPD integration during and after lead optimization

PKPD updated CDTP criteria will impact on the LO phase. It is obvious that a compound's PK characteristics are important and that these are intimately coupled with its in vivo pharmacology and safety margin. It is, therefore, very surprising that emphasis is placed on optimizing PK characteristics in vitro and in vivo and on optimizing target effects in vitro, but not on coupling these to the compound's in vivo properties via a holistic approach (Fig. 1). Optimization of a single property at a

General implications

The principal concern with plasma protein binding is related to its variability within and across different species, strains and disease models in vivo. If we disregard uptake of drug into blood cells, drug in plasma is circulating either as unbound compound Cu or as compound bound Cb to, for example, plasma proteins such as albumin and/or α1-acid-glycoprotein.

The equilibrium processes between unbound and protein bound plasma concentrations and unbound and bound tissue concentrations are

Temporal differences between concentration and response

Different classes of target may have very different temporal relationships between concentration–time and response–time profiles. The anti-thrombotic effect represents rapid equilibria (t1/2 equilibrium less than seconds or minutes, Fig. 7, left) in contrast to the delayed antipsychotic effect of neuroleptics (t1/2 response a few weeks, Fig. 7, right). Thus, the PK requirements for an optimal dynamic effect may vary substantially, depending on the type of target and the target's position in the

Safety/efficacy margins

Because the focus on ‘safety margin thinking’ increases (e.g. handling the hERG/QT issue), it is of utmost importance that complex calculations be based on good science. Essential questions regarding a compound's potency and effective concentration range are connected with determinations of how much of the compound and what duration of exposure are required to produce its desired and undesired effects (onset and intensity) in vivo.

Again, this can be exemplified by test compound A. If the safety

Forerunner information

Clinical (proof of principle, PoP) and preclinical (dog electrophysiological effects) data from test compound B showed that the desired effect was achieved at approximately the same unbound plasma concentrations. Unbound concentrations were, therefore, used as guidance for the effective concentration range of test compound B (Fig. 11). Ideally, one should use the in vivo dog concentration–effect profile to target effective concentrations enabling predictions of human doses.

Clinical information

Dose–effect–time analysis

The most common approach to in vivo PKPD modeling involves sequential analysis of plasma concentration versus time data and effect versus time data, such that the plasma kinetics drives the dynamics. Under certain circumstances, such as for local drug delivery, systemic exposure to drug may not be needed. Hence, systemic concentrations are not measured; nor do they add value. In such a situation, effect versus time data inherently contain useful information about the turnover characteristics of

Study design

Leading into LO studies for test compound B, the plan was to estimate potency for the primary effect (AERP prolongation) and safety margins against unwanted cardiovascular effects in the anaesthetized dog model by repeated rapid intravenous injections of preCD candidates. As a result of close interaction between pharmacology and DMPK, the original protocol was thoroughly revised (Fig. 12).

Major changes included replacement of rapid intravenous injections with a multiple constant drug infusion

Case Study 2: PKPD redesign of in-house safety pharmacology prenomination studies

Safety margins are essential information for different project transitions. A useful example is the safety margins for hERG/QT effects. Generally, a safety margin between the therapeutic (steady-state) plasma level Css and QT prolongation (e.g. in the dog) of a factor of 100 was considered to be necessary to minimize the risk for QT-related arrhythmias in human. As Css is often very difficult to predict (depending, e.g. on target class and compound of interest), such calculated safety margins

Case Study 3: PKPD strategy and dose predictions

This case study exemplifies a situation with large temporal differences between plasma concentration of test compound and the pharmacological response. The LO program for test compound C had an integrated approach to in vivo PK and PD. The parameters governing LO were the in vivo kinetic parameters (CL, bioavailability) and in vivo potency IC50. Test compound C is active at very low total plasma concentrations (fu < 0.1%, high potency), enabling low systemic exposure and a high safety margin. A

Case Study 4: estimating dose of compound D in LO by calibration with forerunner information

The LO program for a back-up to compound D (inhibitor) has combined PK and PD knowledge from preclinical and clinical studies on compound D to identify which back-up compounds have the greatest potential for convenient daily dosing in human. Plasma concentrations of compound D at a known effective dose were simulated based on its known PK. A theoretical relationship between preclinical effect and clinical plasma concentration was then established by coupling the human effective plasma

Case Study 5: modeling of functional adaptation and rebound for the assessment of the impact of slower plasma kinetics

The nicotinic receptor agonist project battles with two fundamental challenges with respect to flushing (adverse effect) and tolerance/rebound of the primary effect (antilipolytic effect, reduction of the plasma nonesterified free fatty acid [NEFA] levels). Nicotinic acid has a low volume of distribution (0.3 L kg−1) and systemic clearance (0.28 L h−1 kg−1), and the bioavailability is good (>80%) in human. It is metabolized to nicotinuric acid (conjugation with glycine; low affinity, high capacity

Case Study 6: linking preclinical disease model information to in vitro screening binding and gene expression data

Compound E was carried through a pivotal chronic in vivo study over a couple of months to establish its disease-curing effect in animals. The outcome was exceptional in that a 45 and 90% disease curing effect was seen at approximately 2 and 20 nm plasma concentrations at steady-state (Fig. 20). Screening data of in vitro binding and gene expression were collated and plotted in the same diagram to show the relationship between in vivo and in vitro characteristics. It is encouraging to note that

Perspectives

Quantitative pharmacological reasoning (PKPD) focuses on concentration–response and response–time relationships with special emphasis on the impact of drugs on disease. The aims of this report are firstly to raise awareness among kineticists and pharmacologists (regardless of their area of therapeutic interest) as to why PKPD integration is essential for TV, optimizing development of lead compounds (LG and LO) and scaling these to human. A related aim is to demonstrate examples of PKPD

Dr Johan Gabrielsson is a Senior Principal Scientist at AstraZeneca R&D Mölndal. He is author of the book ‘Pharmacokinetic and Pharmacodynamic Data Analysis: Concepts and Applications’ 4th ed. (2006). He is academically affiliated with the Department of Pharmacology, Gothenburg University, Sweden. His research focuses on modeling different aspects of endogenous turnover, such as functional tolerance and rebound phenomena by means of feedback, physiological limits and target-mediated drug

References (19)

There are more references available in the full text version of this article.

Cited by (89)

View all citing articles on Scopus

Dr Johan Gabrielsson is a Senior Principal Scientist at AstraZeneca R&D Mölndal. He is author of the book ‘Pharmacokinetic and Pharmacodynamic Data Analysis: Concepts and Applications’ 4th ed. (2006). He is academically affiliated with the Department of Pharmacology, Gothenburg University, Sweden. His research focuses on modeling different aspects of endogenous turnover, such as functional tolerance and rebound phenomena by means of feedback, physiological limits and target-mediated drug disposition in collaboration with Professor Lambertus A. Peletier at Department of Mathematics at Leiden University, the Netherlands. He has conducted numerous workshops on biological (PK/PD) data analysis within and outside the pharmaceutical industry.

View full text