VIEWS & REVIEWS
5
Pharmacokinetics in drug development -Kerry Dechant -
Pharmacokineric/pharmacodynamic modelling Pharmacokinetics is about to enter a new phase of understanding, with emphasis on modelling dynamics using linear and non-linear physiological models: these will enable new developments in drug administration and dosage to evolve. In addition, population dynamics, in conjunction with population kinetics, will improve our understanding of drug activity, disease, and the complexity of physiological processes in the body. These were the conclusions reached by Dr B Campbell (Director of International Scientific Affairs, Servier. Slough. UK) in a presentation on pharmacokinetic/pharmacodynamic modelling at a Management Forum meeting held in London, UK on 24 March 1992. It has now become accepted that pharmacokinetics alone have little relevance without simultaneously measuring the pharmacodynamic activity of a drug. Approaches such as the use of receptor binding models. compartmental models, and physiological models can be used to relate kinetics or blood concentrations to effect. The relationship between drug receptor binding and effect - on a logarithmic plot producing a classical sigmoidal Emax curve profiles the activity of a drug when bound to its receptor [see .figure]. However, this relationship assumes equilibrium between the concentration of a drug in blood and at the receptor site. This is rarely true, particularly during the absorption and distribution phases; thus, the receptor binding model must allow for alterations in blood concentrations, and activity which is out of phase (i .e. hysteresis). Anti-clockwise hysteresis occurs when the curve of drug concentration 1'05 effect (plotted in sequential time order) 'Pharmacokinetics is about to enter a new phase of understanding, with emphasis on modelling dynamics using linear and nonlinear physiological models.' Dr B C.ampbell
progresses in an anti-clockwise direction. It is a reflection of factors such as distribution of drug to its site of action, formation of active metabolites, delayed maximal effects, or sensitisation or increased activity on repeated dosing. Alternatively, clockwise hysteresis (decreasing effect with time for the same drug concentration) can occur because of tolerance, tachyphylaxis or receptor down-regulation; formation of active but antagonistic metabolites; or physiological feedback control regulation. Compartmental dynamic models have been used to counteract problems of distributional
ISSN 0156· 2703/92/0418-0005/$1 ,OO/O@Adj.
""'""lioIlO' lid
hysteresis when steady-state has not yet been reached. These models assume that response is a function of the drug concentration in the central plasma compartment or, more frequently, in a second or third compartment, which cannot be sampled. However, these models oversimplify the real system, and assume that there is a volume of drug at the receptor compartment. A LINK model has been proposed. which makes it possible to measure the lag time for drug action once the drug is in the blood, and to collapse the hysteresis. allowing the use of data even during non-steady-state absorption and distribution phases. Moreover. a recently proposed semiparametric method using devolution enables both clockwise and anti-clockwise hysteresis to be explained. Relationship between drug receptor binding and effect ':r................... . . .. ........'~
{{..
........
1~ 1 40
:
ro
o
,)' , I
·0.7
-0,35
0
0,35
0.7
Log ooncentration (Ko is !he concentration 01 drug required to produce 50% 01 maximal activity)
Recently, however, there has been a shift away from the compartmental approach, as these models have little anatomical basis. and are therefore unreliable and subject to overinterpretation, particularly if the models used are incorrect. Increasingly. physiological models are being developed, which allow a greater mechanistic appreciation of drug activity and how it alters the various body processes involved. Physiological models include the following: • Simple linear model: combines a kinetic model with a series of relatively simple measurable physiological processes, which are often related to rates of synthesis and removal (e.g. effect of warfarin on prothrombin activity through inhibition of vitamin K factors) • Linear system analysis approach: can be used to describe a number of processes which involve physiological feed-back regulation (e.g. control of blood pressure or glucose concentration) • Nonlinear chaos: this model recognises that the body is highly complex and unpredictable, and that any given action can produce many different reactions controlled by various physiological set points. Using the mathematic
INPHARMA® 18 ApI' 1992
VIEWS & REVIEWS principles of deterministic chaos. this model attempts to explain events which, on the surface, appear to be completely random, but underneath have a defined mathematical framework. It can be used to explain biological processes such as cardiac arrhythmias or endocrine control. Dr Campbell concluded by stating that for a complete explanation of kinetic-dynamic relationships, it is clear that nonlinear chaotic theory will be required. In order to totally explain the action of drugs mathematically, all physiological and receptor models will need to be incorporated into a total integrated body systems approach, with each part of the system best described by the best fitting predictive model.
Population pharmacokinetics; theory and practice Population pharmacokinetics describes the design, execution and analysis of pharmacokinetic studies involving sparse data in subjects or patients comprising a particular population of interest. They are important for several reasons: • They can be used to study a population of interest which may differ from the normal population • Pharmacokinetics in certain patients may be altered by pathophysiological factors (e.g. renal disease) • They can be used to quantify and explain variability between patients, which is important for the design of dosage regimens and therapeutic dosage monitoring. Dr L Aarons (Department of Pharmacy, University of Manchester, Manchester, UK) emphasised that the concept of population pharmacokinetics does not differ from what has always been regarded as the primary purpose of pharmacokinetics. Instead, current interest in this field stems from the concern that the pharmacokinetics of new drugs are not studied in relevant populations, i.e. patients likely to receive the drug, at a sufficiently early stage in the drug's development. In a position paper published in 1983, Dr R Temple (Acting Director, New Drug Applications, FDA) recommended that: • Clinical trials should be conducted in target populations and the variability of the responses should be analysed • Drugs should be subjected to a 'pharmacokinetic screen' during premarketing clinical trials. The intention of this proposal was to look for unexpected phenomena and large deviations in important features, such as adverse effects, which could be followed up in subsequent, wellcontrolled studies. Although the optimum time to collect pharmacokinetic data on target populations is during phase III trials, implementation of a
18 Apr 1992INPHARMA~
population approach within the drug development programme is controversial. While its goals are indisputable. i.e. assessment of variability within a population. logistical and ethical reasons preclude intensive experimentation in each patient. Phase III studies are. in general. muIticentre and are frequently performed in an outpatient setting. They can involve large numbers of heterogeneous patients and data are sparse. often limited to a single blood sample per individual. Traditional pharmacokinetic analyses, which rely on extensive sampling in small numbers of homogeneous subjects (frequently volunteers) to determine individual pharmacokinetic parameters, are not feasible. Thus, data analysis techniques which focus on the centra! tendency of pharmacokinetic information, and which are capable of utilising very sparse data are required. Clearly, application of population techniques in fields of medicine which frequently generate sparse data, such as cancer or neonatal medicine, would be highly beneficial. Presently, the approaches utilised for analysis of sparse data include a nonlinear mixed effect model, a non parametric maximum likelihood method, Bayesian methods, and variants of the nonlinear mixed effect model. There is much debate regarding the advantages and disadvantages of each of these methods, usually focusing on issues such as software availability, reliability, and robustness.
'... Population dynamics, in conjunction with population kinetics, will improve our understanding of drug activity, disease, and the complexity of physiological processes in the body.' Dr
n Campbell
Clearly, patient compliance and the accurate timing of both dosing and sampling are critical to the accuracy of population pharmacokinetics. At present, however, there are no guidelines concerning experimental design, in terms of both sample timing and subject numbers, particularly within subgroups. In addition, there is no knowledge of the power of the approach to detect important differences, and there is minimal real experience with the technique. Furthermore, there are no hard data concerning the costbenefit ratio. These are important issues which will need to be addressed. Importantly, Dr Aarons emphasised that population pharmacokinetics is not intended to replace conventional pharmacokinetic studies in drug development. Instead, it is meant to enhance and supplement these data. Traditional pharmacokinetic studies generate excellent data and, as such, they remain necessary for the development of population pharmacokinetics. Nevertheless, a single trough screen allows for
ISSN 0156-2703/92/0418-0006/$1.00/0 © Adi. lnlemational lid
VIEWS & REVIEWS
7
some assessment of variability and analysis of co variates, whereas a multiple trough screen can allow for some separation on intra- and intersubject variability, and a full kinetic screen. with samples spread over the entire dosage schedule, allows for complete kinetic analysis. At present . .... there is too much debate on software and not enough on philosophy. Any data that make statistical sense are useful' concluded Dr
Aarons.
ISSN 0156'2703/92/0418-0007/$1.00/0 © Adi. ~""'-lioooal lid
,,<0
INPHARMA® 18 Apr 1992