Rheumatology Advance Access originally published online on April 26, 2005
Rheumatology 2005 44(7):846-859; doi:10.1093/rheumatology/keh627
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
REVIEW |
Pharmacokinetic–pharmacodynamic correlations and biomarkers in the development of COX-2 inhibitors
1 Division of Pharmacology, Leiden/Amsterdam Center for Drug Research, Leiden University, Einsteinweg 55, P.O. Box 9503, 2300 RA, Leiden, The Netherlands and 2 GlaxoSmithKline, Clinical Pharmacology and Discovery Medicine, UK.
Correspondence to: O. Della Pasqua, Division of Pharmacology, Leiden/Amsterdam Center for Drug Research, Leiden University, Einsteinweg 55, P.O. Box 9502, 2300 RA Leiden, The Netherlands. E-mail: odp72514{at}gsk.com
 |
Abstract |
|---|
Top
Abstract
IntroductionThe mechanism by which COX inhibitors exert their analgesic effect is well established. However, data show no direct correlation between drug concentrations in plasma and the analgesic or adverse effects in chronic inflammatory conditions. This represents a major problem in the development of COX inhibitors, since it is difficult to predict the appropriate dosing regimen for the treatment of chronic inflammatory pain, based upon information from pre-clinical studies and eventually early clinical studies. The factors that determine response in inflammatory pain must be understood in order to make predictions about the time course of the analgesic effect. In this review the determinants of drug response and their variability will be discussed: physicochemical properties, pharmacokinetics (PK), pathophysiology and disease progression.
From a mechanistic point of view, endogenous mediators of inflammation might be used as a biomarker for the analgesic effect and safety assessment. Such a biomarker can be an intermediate step between drug exposure and response. In addition, its concentration–effect relationship could be characterized by pharmacokinetic–pharmacodynamic (PK/PD) modelling. Indeed, recent investigations have shown that COX-2 inhibition, as determined by modelling of prostaglandin E2 (PGE2) levels in the whole blood assay in vitro can be used as a marker to predict drug effects (analgesia) in humans. A model-derived parameter, IC80, (total and unbound) was found to correlate directly with the analgesic plasma concentration of different COX inhibitors varying in enzyme selectivity. These findings indicate that PGE2 and thromboxane B2 inhibition can be used to predict and select efficacious doses in humans.
|
Introduction |
|---|
When the outcome of pre-clinical or clinical experiments is very difficult to quantify and shows large variability, as observed in pain measurements, a biomarker of drug effect might be useful [1]. Biomarkers have been defined by the Biomarkers Definitions Working Group as physical signs or laboratory measurements that may be detected in association with a pathological process and that have putative diagnostic and/or prognostic utility [2]. A biomarker that is intended to substitute for a clinical endpoint is a surrogate endpoint. Biomarkers can be evaluated in different models, i.e. in vitro assays, in vivo animal models, clinical trials in healthy subjects and patients, epidemiological studies and simulated biological systems.
During the early stages of drug development, biomarkers can be used as a guide for dose selection, dose escalation and safety monitoring. In diseases in which the pathogenic mechanism is poorly understood, biomarkers might also be useful. In fact, biomarkers can play a crucial role in understanding on a mechanistic level the differences in clinical response between subjects [3]. Biomarker measurements can also help explain the empirical results of clinical trials by relating the effect of interventions on molecular and cellular pathways to clinical response.
The validation of biomarkers is, therefore, very important, particularly if one bears in mind their relevance for decision-making and regulatory purposes. Their use as surrogate endpoints is evaluated by considering those factors that are related to the ability to accurately substitute for a clinical endpoint, including face, criterion and construct validity [4]. In addition, practicality, specificity and innovation are three important aspects in the development of biomarkers or surrogate endpoints. Typically, for innovative products limited information is available on the relationship between biomarker versus clinical response. This may limit the utility of biomarkers in the early development phases. In contrast, the availability of a validated biomarker may greatly facilitate the development of compounds that act via well-established mechanisms of action [4].
There is growing evidence that the evaluation and validation of biomarkers according to mechanism-based pharmacokinetic–pharmacodynamic (PK/PD) modelling and simulation can increase the predictive power of biomarkers and their use in drug development.
A seven-point mechanistic classification based on the location of the biomarker in the chain of events, from the underlying subject's genotype or phenotype through to clinical scales was proposed by M. Danhof at the Ninth European Federation for Pharmaceutical Sciences (EUFEPS) Conference on Optimizing Drug Development held in Basel, Switzerland, on 10–12 December 2001 [2]. This mechanism-based classification consists of distal to proximal to clinical endpoints. The different types of biomarkers are type 0 biomarkers (genotype or phenotype), type 1 biomarkers (concentration), type 2 biomarkers (target occupancy), type 3 biomarkers (target activation), type 4 biomarkers (physiological measures or laboratory tests), type 5 biomarkers (disease processes) and type 6 biomarkers (clinical scales).
In this review first we discuss the determinants of treatment response to COX inhibitors and the role of PK/PD modelling in the development of COX inhibitors. Secondly, the role of biomarkers to correlate the non-linear relationship between analgesia and exposure, as measured in different clinical and pre-clinical models, will also be evaluated. We will focus on the pre-clinical in vitro and in vivo models and subsequent scaling of concentration–effect relationships from pre-clinical to clinical settings.
|
COX inhibitors |
|---|
COX inhibitors are effective anti-inflammatory, antipyretic and analgesic agents which are commonly used in the treatment of acute and chronic pain, rheumatoid arthritis and osteoarthritis. They act by inhibiting cyclo-oxygenase (COX) activity and consequently the formation of pro-inflammatory mediators like prostaglandins (PG) and thromboxanes (TXB) (Fig. 1) [5]. Since the early 1990s, it has generally been accepted that cyclo-oxygenase exists in two isoforms. Cyclo-oxygenase-1 (COX-1) is a housekeeping enzyme responsible for modulating physiological events and is present in most tissues including stomach, kidney and platelets, whereas cyclo-oxygenase-2 (COX-2) is highly induced in various cells by pro-inflammatory stimuli, mitogens and cytokines [6]. Recently, a third isoform (COX-1v) has been identified, but its function is not yet clear. Continuous COX-1 inhibition is thought to be principally responsible for gastrointestinal adverse effects following prolonged administration of non-selective COX inhibitors, whereas selective COX-2 inhibition accounts for the anti-inflammatory, anti-pyretic and analgesic efficacy [7]. Recent investigations have demonstrated that the roles of COX-1 and COX-2 are oversimplified. Data from those studies suggest that COX-2 is present under non-pathological conditions in tissues such as kidney, brain and the spinal cord, playing an important role in the maintenance of physiological homeostasis [8]. Previous findings have also shown that selective inhibition of COX-2 only partially reduces the level of PG at the site of inflammation in comparison with non-selective COX inhibitors, which reduce PG to undetectable levels [9]. This suggests that COX-1 may contribute to the pool of PG at the site of inflammation. Yet, COX-2 does seem to have emerged as the dominant (albeit not exclusive) source of PG formation during inflammation [10].
|
By reducing PG and TXB synthesis, COX inhibitors block the nociceptive response to endogenous mediators of inflammation, with the effect being greatest in tissues that have been subjected to injury or trauma (Fig. 1). COX inhibitors also exert their analgesic effects through a central mechanism. Central neuronal tissues have been shown to synthesize PGs, and spinally administered COX inhibitors have been shown to reduce peripherally induced hyperalgesia [11].
Rational drug therapy is based on the assumption that there is a causal relationship between dose, dosing regimen and/or exposure to a medication on one hand and its therapeutic response as well as adverse effects on the other. Hence, it has been one of the major goals of clinical pharmacology to find systematic ways to identify the dosing regimen that produces clinically relevant analgesia [12]. An important question that remains to be answered is how much and for how long COX-2 and COX-1 should be inhibited to ensure an optimal risk–benefit ratio, allowing for a sustained analgesic response and appropriate safety margin. To date, the dose selection of COX inhibitors has been based primarily on clinical endpoints for analgesia, an approach which disregards the impact of maximum, long-lasting blockade of either enzyme systems.
Despite a clear analgesic response upon administration of COX inhibitors, no direct correlation is observed between the drug concentrations in systemic fluids (blood, plasma) and the magnitude of the pharmacological effect. The lack of such a direct correlation does not mean that one cannot identify the optimal dose required to achieve a pre-defined treatment response. Adequate insight into the pathophysiology of pain in conjunction with tools such as non-linear mixed effect modelling can enable the characterization of non-linear relationships between pharmacokinetics and pharmacodynamics.
|
Complex signal transduction processes of pain |
|---|
Understanding of the processes and mechanisms underlying pain arousal and transduction is essential to establishing a link between analgesic effect and exposure to anti-inflammatory drugs. However, most attempts to correlate the analgesic effect with drug exposure have been based on absolute and relative scales of pain perception or relief, instead of assessing the rate and extent of change in mediators responsible for pain signalling and transduction. Such measures are independent of numerous external factors that can affect pain perception and pain relief. In addition, high variability in pain measurement in animals and humans, especially when compared with other sensory systems, has been reported in an abundance of articles. Progress has been made in the understanding of pain physiology, but pain perception in animals and humans nonetheless displays considerable and unexplained inter-individual variability [13]. Moreover, inter-individual variability greatly exceeds intra-individual variability in pain measurements [13].
Physiological pain occurs when electrical signals are transmitted via A-mechanothermal and C-polymodal fibres to the thalamus and higher centres of the brain (Fig. 2). The afferent nociceptors are peripheral nerve endings, with a high threshold for activation [14]. In inflammatory pain, hypersensitivity is the consequence of alterations in transduction sensitivity of the nociceptors, activity-dependent changes in excitability of spinal neurons and phenotypic changes in sensory neurons innervating the inflamed tissue [15]. Tissue injury results in the release of inflammatory mediators from damaged or infected cells, increasing the transduction of painful stimuli. Among these mediators, a second series of signals is that generated or initiated by inflammatory cytokines such as tumour necrosis factor (TNF-
) and interleukins (IL-1ß, IL-6). Cytokines act on and between inflammatory cells, inducing some of the features of the inflammatory response. They also mediate some of the systemic effects of inflammation, such as fever or cachexia [15]. The induction of cytokines can lead to the expression of the inducible form of nitric oxide synthase (iNOS), which in turn provokes the release of excessive amounts of nitric oxide (NO) that may participate in the pathogenesis of tissue injury. However, the interplay between different cytokines and inflammatory mediators like prostaglandins is not fully understood yet [16]. In addition to local cellular events, potassium, prostaglandins, bradykinins, ATP and other mediators from damaged cells trigger the nociceptors to send afferent impulses via the dorsal root ganglion to the spinal cord. Afferent information is transmitted via second-order neurons in the dorsal horn through the spinothalamic tract to the thalamus and to the sensory cortex [17]. These higher centres are responsible for the perception of pain. At different levels of the pain pathway, a complex system of compensatory neuroinhibitory mechanisms is involved in the perception of pain.
|
The inflammatory process is a dynamic process with varying phases depending on the status of the disease. Typically, the release of chemical mediators is variable in onset and duration. In pre-clinical experiments in rats, concentration–time profiles of a spectrum of chemical mediators released in peripheral tissue, plasma and the central nervous system after intraplantar injection of Freund's complete adjuvant (FCA) show wide variability (Table 1). Variability in the time to maximum concentration of different mediators after FCA injection in Table 1 ranges from within a few hours to several weeks after inflammation. Furthermore, increased production of mediators may last from one to several days. Interestingly, the time course of the elevation of certain mediators may also show biphasic behaviour. For example TNF-
shows a peak level 24 h after FCA injection followed by a decrease to baseline levels until the TNF- level rises 5 days after FCA injection. This might indicate a dual role for TNF- in the onset and maintenance of the inflammatory response.
|
In addition to fluctuations in chemical mediators, measurements of pain (algesia) reveal large variability over time. The analgesic effects of COX inhibitors have been studied in a wide range of animal models. The main findings of these investigations are summarized in Table 2 In this table, the effects of Ca2+-channel blockers and opioids are shown, for comparative purposes. Comparison of drug effects reveals large differences in the ED50 values, the dose that produces 50% of the maximum effect (Table 2). For example, aspirin ED50 values vary between 51.8 and >200 mg/kg for the different animal models of inflammatory pain. Since it is unclear whether animal models truly reflect inflammatory disease states in humans, it is not adequate to make direct extrapolations from the efficacious dose in the FCA model to predict therapeutic dose in human rheumatoid arthritis. Moreover, variability in ED50 could be due to differences in PK between different animal models, which are not taken into account when the ED50 parameter is compared. A more scientific approach should consider the assessment of drug exposure, i.e. EC50, the concentration that produces 50% of the maximum effect, rather than ED50. However, large variability in analgesic effects and algesia exists in both animal and human data, which makes it difficult even to compare the different concentration–effect relationships of COX inhibitors. Scaling and parameterization of the data based on a biomarker, which is likely to show less variability itself, might enable mechanism-based extrapolation between the different models and species. In addition, biomarker analysis in combination with pain measurements might help to understand variability in pain perception.
|
|
Pharmacokinetic–pharmacodynamic modelling of COX inhibitors |
|---|
The primary objective of PK/PD modelling is the characterization and prediction of the time course of drug effects under physiological and pathological conditions in vivo [18]. Such PK/PD models consist of (i) a pharmacokinetic model, describing the time course of the drug in blood or plasma, (ii) a pharmacodynamic model, describing the relationship between drug concentration and effect and (iii) a link model, which serves to account for the often observed ‘delay’ of the effect relative to the plasma concentration [18].
A limitation of the currently available PK/PD models for COX inhibitors is that they are descriptive and do not take into account the pathophysiology of inflammation and the role of disease progression. Specifically, changes in disease status as a function of time (e.g. as reflected by varying concentrations of endogenous mediators) and the mechanism of COX inhibition need to be taken into account in order to fully understand the non-linearity in the relationship between PK and analgesic effects during acute and chronic inflammation (Fig. 3).
|
|
Measurement of pain |
|---|
Animal models of pain
Thus far, it has been assumed that the pre-clinical models that mimic clinical inflammation in humans can be used to predict the analgesic effect of COX inhibitors. According to current understanding of the underlying mechanisms, pain models can be broadly classified into three types: (i) physiological (acute nociception), (ii) inflammatory and (iii) neuropathic pain models [19]. Application of noxious stimuli results in an increased pain response in models of neuropathic, inflammatory and post-operative pain syndromes compared with the healthy situation, which is called hyperalgesia. Pain evoked by an application of innocuous stimuli is called allodynia.
In models of physiological pain, a noxious stimulus, mostly noxious heat, is used to measure the behavioural response of naïve animals [20]. In the hotplate and tail flick tests, the paw or tail is heated by a radiant heat source and the latency time to withdrawal is measured [21]. These models can be used to characterize the pharmacological effects of analgesic and antihyperalgesic drugs. However, they have the disadvantage that drug administration occurs prior to heat stimulation. In general, COX inhibitors do not affect the normal nociceptive response threshold or latency time in the hotplate or tail flick tests (Table 2).
Models of inflammatory pain produce a state of hyperalgesia that is evident from a few hours to weeks after peripheral injection of chemical irritants [21]. Alternatively, inflammatory pain can also be induced by exposure of the rat hind paw to UV irradiation, which produces highly reliable and persistent allodynia [20]. Similarly to the physiological models mentioned previously, drugs are often administered pre-emptively, i.e. before application of pain stimuli. Mechanical hyperalgesia due to inflammatory pain can be reduced in a dose-dependent manner by COX inhibitors (Table 2) [22].
In most of the mononeuropathy models sciatic nerve injury is produced which leads to abnormal pain in the hind paw. Today, three different models of neuropathic pain are most frequently used: (i) the chronic constriction injury (CCI) model as proposed by Bennett and Xie [23]; (ii) the single tight ligation of the partial sciatic nerve (PSL) model by Seltzer et al. [24] and (iii) the segmental spinal nerve ligation (SNL) model by Kim and Chung [25]. In these models, hyperalgesia and allodynia typically develop over several weeks. However, the onset and extent of each component is different between the models. Only a few articles have been published showing that COX inhibitors attenuate hyperalgesia in models of neuropathic pain [26].
Correlation between the potency of COX inhibitors in animals and humans
Although animal models of inflammatory and neuropathic pain have similarities to human disease conditions, extrapolation from pre-clinical animal models is rather difficult. Dubinsky et al. [27] observed a correlation between the average single oral human therapeutic dose for pain relief and the oral ED50 dose in the joint flexion assay in arthritic rats. Mukherjee et al. [22] observed a correlation between the ED50 for paw oedema in the carrageenan model and the effective oral human daily dose for an array of COX inhibitors. These results indicate that information from pre-clinical pharmacodynamic models can be of value for the prediction of the analgesic dose of COX inhibitors in humans.
In the analyses of both Dubinsky et al. and Mukherjee et al. a linear relationship was observed according to:
 | (1) |
|
Models to investigate selectivity on COX-1 and COX-2 |
|---|
As an alternative to in vivo measurements, in vitro methods may enable the assessment of concentration–effect relationships that reflect in a strict quantitative manner the anti-inflammatory response in vivo.
Different models have been developed to study the selectivity of COX inhibitors on cyclo-oxygenase activity in vitro [28]. These models include purified or microsomal recombinant enzymes [29, 30], homogenized cells [31, 32] or intact cells [33, 34]. The advantages and drawbacks of each individual test have been analysed elsewhere [35] and will not be discussed in such detail in this review. As standardization is not available, experimental conditions can vary widely among the various assays. In these assays arachidonic acid can be supplied from endogenous or exogenous sources, which can affect the apparent potency of competitive (Class 1) COX inhibitors. It has been demonstrated that these models can be useful to guide the synthesis of selective COX-2 inhibitors [28], but such models do not consider protein binding of COX inhibitors and other aspects of the kinetics of drug–enzyme interaction [6]. In fact, some sources of variability have been identified, like time-dependent binding for certain compounds, which could be adapted when investigating different classes of COX inhibitors. Therefore, assay systems that reflect physiological conditions and account for differences in pharmacological properties provide the most appropriate models [36].
A more pathophysiological model to study the selectivity on COX-1 and COX-2 is the human whole blood assay (hWBA), which was first described by Patrignani et al. [28]. In the hWBA, the production of prostaglandin E2 (PGE2) by bacterial lipopolysaccharide (LPS) stimulation in leucocytes in the whole blood is used to characterize drug effects on COX-2, whereas the production of thromboxane B2 (TXB2) by platelets in the whole blood sample during 1 h whole blood clotting is used as an endpoint to determine drug effects on COX-1. The hWBA has several advantages over other in vitro systems, as it uses whole cells, takes into account intra-cellular transport of drugs and provides a physiological plasma protein level. Moreover, the method can be used ex vivo following in vivo administration of COX inhibitors at therapeutic doses [37]. A reflection of therapeutic efficacy is achieved when the hWBA is used to determine the effects of COX inhibitors in healthy volunteers who have taken the drug for several weeks [36]. Moreover, this model enables calculation of IC50 values (i.e. the concentration that produces 50% of the maximal inhibitory response). A drawback of this system is still the high biological variability in measurements, but that does not surpass its advantages as long as sample size is properly considered [37]. Blain et al. [37] found that the level of COX-1 and COX-2 inhibition achieved in vivo cannot be predicted universally from the corresponding in vitro dose–response curves, which could be a major drawback of the model. As they state, the predictability of the hWBA was suitable for diclofenac, acceptable for ibuprofen, but inadequate for meloxicam. However, the investigators only plotted the average concentration–inhibition curves of 24 healthy volunteers without evaluating information derived from between-subject variability. Non-linear mixed-effects modelling of individual concentration–inhibition curves as proposed in this review was not applied to that data set.
The estimated IC50 values for COX-1 and COX-2 of different COX inhibitors have been shown to vary between models. Yet, the ranking order of selectivity remains approximately the same from one study to another. COX inhibitors with the highest selectivity for COX-1 and the lowest for COX-2 and those with the lowest selectivity for COX-1 and the highest for COX-2 yield the same results, irrespective of the preparation used for the assay [38]. In addition, the hWBA has been used for the determination of COX-1 and COX-2 selectivity in human, rat and guinea pig blood [28, 39, 40]. Despite methodological variations, when comparing all the different assay systems, the COX-2/COX-1 ratios of different COX inhibitors are able to classify the COX inhibitors according to their selectivity into selective COX-2, non-selective and selective COX-1 inhibitors [41]. However, the assessment of COX-1/COX-2 selectivity ratio ex vivo after drug administration to humans has provided evidence that selectivity behaves as a continuous variable. Thus, it will reflect the selectivity in vitro, but more importantly the dose and frequency of administration. This is further amplified by an important intra-subject variability in both COX-1 and COX-2 inhibition after dosing.
|
Selection of COX inhibitors |
|---|
In the early phases of drug development, the selection criteria for new COX inhibitors are primarily based on the inhibition profile for COX-1 and COX-2. Typically, drug selectivity is assessed by in vitro assays during screening, as these assays are relatively rapid and simple. In these models, the selectivity of action of COX inhibitors is determined by the ratio of IC50 estimates for the inhibition of COX-1 and COX-2, respectively. Comparison of the IC50 ratio is frequently used to determine the relative biochemical selectivity of the various compounds [42]. Despite the widespread use of these assays for screening purposes, translation of the results from selectivity profiles to the analgesic effect in vivo has not been reported.
To date, the dose selection for COX inhibitors is targeted to maximal inhibition of COX-2. Typically, no attention is given to the dose range at which the therapeutic effect occurs without loss of selectivity of action, or the minimal level of inhibition required to produce a therapeutic effect. This approach may lead to an undetectable increase in safety risks. As depicted in Fig. 4C (adapted from [42]), a typical COX inhibitor has an in vitro COX-1/COX-2 ratio of about 18, which indicates that the compound can be classified as a selective COX-2 inhibitor. At a single therapeutic dose of 50 mg or 100 mg, however (Fig. 4A and B), the compound causes a significant inhibition of COX-1 at drug levels that suppress COX-2 by >90%. This shows that this drug cannot be considered a preferential COX-2 inhibitor after administration of ‘therapeutic’ doses [42]. Figure 4C shows that at a concentration resulting in 90% inhibition of COX-2, COX-1 is inhibited by at least 40%, which accurately reflects the drug profile in vivo. In this respect it is important to realize that the use of an in vitro ratio based on IC50 values does not capture differences in the shape of the concentration–inhibition curves for COX-1 and COX-2, which can be very different, as indicated by different slopes. As a consequence, the actual selectivity ratio varies with the concentration, making it incorrect to determine the selectivity of action on COX-1 and COX-2 solely based on the ratio of IC50 values. In fact, this ratio should only be used as an approximation of selectivity in the absence of information on the slope of the concentration–inhibition curve. It is not yet understood which mechanisms determine the slope of the drug concentration–inhibition relationship.
|
Table 3 shows the dosing range of different COX-inhibitors, IC50 values for the inhibition of COX-1 and COX-2 in vitro (both in µM), and the therapeutic plasma concentration in humans. The therapeutic plasma concentration in µM is the average concentration in plasma at which analgesia is achieved in patients. Typically at the therapeutic analgesic dose, COX-1 is inhibited by more than 80%, except for COX inhibitors that display more than 1.56-fold selectivity for COX-2. This demonstrates that an in vitro IC50 ratio can predict the inhibition in vivo. If a compound is more than 1.56-fold selective for COX-2 in vitro (as assessed by the IC50 ratio), only COX-2 will be inhibited in vivo at therapeutic plasma concentrations. Although this relationship has been observed between in vitro and in vivo COX-1 and COX-2 inhibition, no direct correlation has been established for analgesic effects of COX inhibitors. Understanding of other relevant factors is required to enable early, accurate prediction of therapeutic doses.
|
|
Determinants of the PK/PD relationship of COX inhibitors in acute and chronic inflammatory pain |
|---|
One may need to consider more than just the pharmacokinetics and pharmacodynamics to determine time course of the treatment response of COX inhibitors in inflammatory pain. Specifically, time-dependent factors in the pharmacodynamics of enzymatic inhibition and the role of disease progression must be taken into account.
|
Pharmacokinetics |
|---|
The physicochemical properties of a drug contribute to its pharmacokinetics. In general, COX inhibitors are lipid-soluble, weakly acid, drugs which bind extensively to plasma proteins, with albumin as the major binding protein [43]. Typically, COX inhibitors have a low extraction ratio and are subject to restrictive elimination (i.e. only clearance of the unbound fraction occurs). Thus, changes in plasma protein binding can have significant effect on drug disposition, volume of distribution (Vd), clearance (Cl) and pharmacological activity [43]. Therefore, the unbound fraction of a drug, fu, is a critical parameter in describing and understanding the pharmacokinetics of COX inhibitors.
The pKa, the acid dissociation constant, of COX inhibitors is another important determinant of treatment response, as it drives the absorption rate after oral administration. In reality, it has been hypothesized that the pKa value can be an important determinant of the distribution of COX inhibitors. It appears that un-ionized COX inhibitor concentrations in the extracellular fluid increase with increasing pKa values, which allows enhanced diffusion of drug to the site of action [22]. Generally, COX inhibitors have a relatively small volume of distribution, which indicates preferential distribution to certain body compartments. Selective tissue accumulation of COX inhibitors at the site of inflammation may explain their superior anti-inflammatory analgesic activity, but is also known to cause gastrointestinal and kidney toxicity [44].
The lipid solubility of drugs is another determinant of the ease with which drugs cross cellular membranes. The octanol–water partition coefficient, determined at pH 7.4, could be the best indicator for the transmembranal diffusion of COX inhibitors. On the other hand, Mukherjee et al. [22] found that the octanol–water partition coefficient at pH 7.4, and therefore the diffusion to intracellular sites, is not a limiting factor for this class of drugs.
|
Equilibration kinetics |
|---|
Physicochemical properties contribute to the affinity and efficacy of COX inhibitors. To date, it is generally accepted that COX inhibitors can be classified on their mode of inhibition of COX in three classes: Class 1 inhibitors are simple competitive inhibitors, Class 2 COX inhibitors are time-dependent inhibitors and Class 3 COX inhibitors are irreversible inhibitors [38]. In contrast to classical non-selective COX inhibitors, which are Class 1, 2 or 3 for both COX enzymes, all diarylheterocyclic COX-2 selective inhibitors like celecoxib and rofecoxib are Class 2 for COX-2 and Class 1 for COX-1. Thus, selective COX-2 inhibitors are Class 1 inhibitors for COX-1, binding to COX-1 for a short period, but are Class 2 inhibitors for COX-2, and therefore display time-dependent binding to COX-2. On the other hand, it has been demonstrated that drugs like nimesulide and meloxicam, which have time-dependent binding to COX-2, also display time-dependent binding to COX-1 [30]. However, these classical COX inhibitors bind differently to the enzyme than the newly developed coxibs.
Class 3 inhibitors irreversibly inhibit COX. Aspirin is the only example of a marketed class 3 inhibitor. The pharmacodynamics of aspirin depends on the synthesis and degradation of the COX-enzyme. Yamamoto found that after administration of a single dose (325 mg) of aspirin to healthy volunteers, aspirin concentrations were measurable for 8 hours in plasma, whereas COX-1 was inhibited for approximately 200 hours [45]. This clearly shows that the turnover of COX can be a major determinant of the pharmacodynamics of COX-inhibitors. To what extent such time-dependencies influence the pharmacodynamics of COX-inhibitors is not yet clear, but obviously it depends on the relationship between drug clearance, distribution and offset (koff) of the binding to COX. For class 1 COX-inhibitors, no influence of COX turnover is expected since the binding is more rapid than the elimination half-life.
Gierse hypothesized that determination of the selectivity based on the ratio of IC50 values is not supportable if the underlying binding kinetics of COX-inhibitors is different for each isoform of the COX-protein. This underscores the need to characterize the in vivo selectivity of action under physiologically relevant conditions [46].
|
Disease progression |
|---|
Disease progression can be defined as changes in disease state as a function of time [47]. Drug effects may be symptomatic (i.e. no change in the disease progression), protective (i.e. the drug causes a slower rate of disease progression) or curative (i.e. the drug abolishes disease progression).
Disease progression analysis describes in a strict quantitative manner the effect of drug treatment on the progression of disease. This is important in the following situations: (i) when the drug has no direct observable pharmacological effect, (ii) when the drug is intended to modify disease progression and (iii) when disease severity influences the drug response.
The dynamics of disease progression determine to what extent a drug can yield a pharmacological response (Fig. 5). The absolute acute effect of a drug depends on the status of the disease. In Fig. 5, at time T, an absolute acute effect of A can be reached, whereas at time T + 1, a maximal effect of B can be reached. In order to predict the effect of a drug in a chronic disease, the time course of disease must be characterized. In addition, if a drug is given chronically, a separation must be made between symptomatic and disease-modifying effects of the treatment. In contrast to opioids, the analgesic effects of COX inhibitors have been primarily studied under chronic disease conditions because COX inhibitors have no hypoalgesic effects. However, this does not preclude the assessment of underlying mediators of inflammation, which could be performed in animals and humans. Melli [48] investigated the levels of PGE2, a marker for COX-2 activity, in chronic inflammatory processes in rats. The plasma levels of PGE2 showed oscillations over time, indicating that the amount of COX-2 protein is not constant or that the metabolic rate varies under disease conditions. These findings suggest that the effects of COX inhibitors in the treatment of chronic inflammatory conditions may be variable, if inhibition levels are not optimal. Consequently, the effects of COX inhibitors upon the chronic inflammatory processes could be different from those observed in acute conditions.
|
|
Biomarkers |
|---|
The nature and complexity of the interaction between various factors that determine the analgesic response of COX inhibitors require biomarkers to explain and understand variability in the treatment effect. The use of a biomarker in pain measurements is an important step in the development of new COX inhibitors, as it can link PK to the analgesic effect and eventually provide a proxy for safety evaluation. Given the nature of the inflammatory response and the mechanism of action of COX inhibitors, a number of mediators can be used as an intermediate step between PK and analgesia. In conjunction with non-linear mixed-effect modelling, the relationship between biological marker, pain measurement and safety can be characterized. Primary candidates for such a role are prostaglandins and thromboxanes (Fig. 1). The mechanism of inflammation in animals and humans is similar, which might facilitate the extrapolation of this biomarker from pre-clinical to clinical data. In the next subsection, literature findings of biomarkers and correlation with pain response are presented. Pre-clinical data are limited, and therefore the next step is to use clinical data to search for a suitable biomarker to predict analgesic response.
|
Relationship between biomarker and pain |
|---|
Limited information is available on the time course of the intensity of pain and biomarkers in animal and human studies. Figure 6 shows the correlation between pain response and spinal reorganization over time following FCA injection. The maximal increase in neurokinin-1 receptor density is correlated with the maximal pain response 4 days after FCA injection. The same correlation between pain response and receptor density is observed in the CCI model [49]. These results indicate that pain response can be linked to a specific substrate in chronic diseases in animals. However, the use of receptor occupancy or tissue samples in the context of drug development is laborious and a systemic biomarker might be preferable. A systemic biomarker, however, should reflect the underlying processes in the target tissue or a mechanistic extrapolation between target tissue and systemic exposure has to be demonstrated. To date, no literature findings of systemic biomarkers for pain measurements are available from animal studies, and therefore we proceeded with clinical data. Cryer and Feldman [50] found a direct correlation between the gastric IC50 and the COX-1 IC50 in blood, as measured by the hWBA of both selective and non-selective COX inhibitors in humans. In biopsies, total gastric COX activity was measured by PGE2 production. Inhibition of COX-1 can be used to predict the gastric dose; however, it is not clear whether this gastric dose truly reflects side-effects like ulcer formation. Still, this is a first step in the development of a systemic biomarker for pain measurements.
|
