PKPD modeling integrates a pharmacokinetic and a pharmacodynamic model component into one set of mathematical, statistical and numerical expressions that allows the description of the time course of effect intensity in response to administration of a drug dose [1]
A major goal in clinical pharmacology is the quantitative prediction of drug effects. The field of PKPD-modeling has made many advances from the basic concept of the dose-response relationship to extended mechanism-based models. The first response relationship from the first classic models developed in the mid-1960s to some of the more sophisticated current approaches (NONMEM).[ ]
Use of modelling and simulations must be accompanied by the critical infrastructure elements to be successfully deployed in drug development. Modelling requires specialised software and experienced analysts. It is generally felt that there are not enough experienced analysts in the field at the moment and that more training is needed (2). Critical infrastructure needs to be developed to enable efficient pooling of data across trials and across programs.
PKPD Modelling generally can add value in all stages of drug development. Its benefit and value will depend on the amount of data already existing for the particular active pharmaceutical ingredient (API) or the data for related compounds (same structural class, class effects. To have the optimal use and benefit of modelling and simulatation, PKPD modelling should be established at the very earliest drug development stage or phase, preferably during the preclinical phase.
These models can be updated continously, when more and more data from later phases becomes available. Their validation is necessary during development, and they will then provide valuable support to make important decisions, with an increased conï¬dence level around the analysed data [13,14].
Modelling here is often based on assumptions such as the relative efï¬cacy and potency in animal efï¬cacy models between the new compound, and comparators are predictive of relative efï¬cacy and potency in humans and that allometric scaling gives a reasonable estimate of the clearance in humans [16,17]. PK/PD models should include the rate-limiting component from the mechanism of action (e.g. receptor binding, protein synthesis). Other covariates, such as protein binding, receptor occupancy (species difference), active metabolites and competitive endogenous ligands can all affect a model's predictivity and need to be considered, if possible. For conï¬dentiality reasons, published examples of modeling data for compounds in preclinical and early clinical development are rare. A recent example is the publication on the in vivo characterisation
blood-pressure-lowering drug. On the basis of the potency data (EC50), the newcompoundwas expected to be less potent than the comparator. However, the maximal effect model (Emax), built using the PK and efï¬cacy data from animal models and human data for the comparator, predicted that the new compound would have a higher Emax, suggesting a superior efï¬cacy at higher con-centrations [14]. Thus, a compound initially seen as non-competitive was, as the modelling predicted, of superior efï¬cacy in the clinic. The greatest cost saving potential of modelling in preclinical phase is allowing the selection of the optimal drug candidate and abandoning early those which are not predicted to exhibit required efï¬cacy or safety [8].
Beneï¬ts of modelling in early clinical development (Phase I)
Phase I
Phase I studies provide initial human data for the tested compound and include a small number of short studies in healthy subjects or patients. They provide early data on human tolerability, PK and sometimes PD.Owing to their short duration and small size, clinical efï¬cacy endpoints can rarely be tested in these studies.
A model developed during the preclinical phase can be further reï¬ned to incorporate early clinical data for the compound [19]. Modelling of early clinical data can sometimes provide invaluable information as in some of the cases below:
- If PK is non-linear, only modelling can help in estimation of relevant parameters and concentration/time proï¬les.
- In complex relationships between PK and PD, only modelling can describe the course of the PD in question.
- If sparse sampling is the only option (e.g. in a paediatric population), modelling is the only option for an adequate and efï¬cient interpretation of the data.
- A PD model based on preclinical or biomarker data, or a PK model based on Phase I data, can be valuable for designing Phase II trials. This is especially useful for compounds with a known potency that can be compared with a drug with a known
response in patients.
- Modelling of the multiple dose data can be used to identify enzyme auto-induction or the development of tolerance.
- Modelling based on in vitro dissolution/in vivo absorption relationship may reduce the in vivo experimental workload for formulation testing.
- Modelling to compare merits of sparse and intensive blood sampling may help to select the optimal number and time points [2,19]
Population PK modelling, based on Phase I data, should be pooled across multiple studies. However, the expected PK variability in patient population and speciï¬c covariates cannot be reliably identiï¬ed in Phase I studies. Yet, the information obtained using such preliminarymodel of Phase I data togetherwith protein
binding and in vitro efï¬cacy data is needed to enable selection of the most efï¬cacious dosing regimen and sampling for Phase II. Early modelling can also enhance conï¬dence in predetermined crucial success factors, aiding the decision making process.
Beneï¬ts of modelling in medium clinical development (Phase II)
Phase II includes studies in carefully selected patients from the patient population of interest. They provide data across a dose range and help to assess a dose-response relationship. Customarily, Phase II consists of a fairly large number of studies, which run in parallel or sequentially, in order to provide sufï¬cient information to guide Phase III planning. PK/PDmodelling and simulations can signiï¬cantly reduce the number of Phase II studies needed to provide the data required for further development. Modelling and simulation in Phase II can be used to
- Develop a drug-disease model to understand the time course of disease progression and dose-response to interventions.
- Test different study designs (including ones that are not customarily used) and assist in selection of the optimal design for the given conditions.
- Assess the probability of success, given a study design.
- Design optimal dosing and sampling schemes.
- Simulate outcome, given assumptions and study design considerations.
- Assess the impact of covariates, using a population PK/PDmodel.
- Assess the efï¬cacy/toxicity proï¬le, relative to comparators [2,15,20,21].
PK/PD modelling in early clinical development may enable crucial decisions (e.g. go/no go) to be reached earlier, based on PK and PD characteristics of the compound, especially if models include comparators' data. Abandoning compounds with suboptimal PK/PD characteristics early in their development leads to signiï¬cant cost reduction [20]. Conversely, by applying a baseline-response model (e.g. adjusted for disease severity), it is sometime possible to discern desired efï¬cacy in compounds initially considered as non-competitive [2].With a good time-response model it is possible to estimate steady-state effects and predict results fo
varying study durations and optimal timing of the study visits. Modelling and simulation of the data from only a few well designed studies can provide sufï¬cient information that would otherwise come from a larger number of separate studies [2,19]
This can potentially provide a signiï¬cant saving by reducing the totalnumberof Phase II studiesneeded.Additionally, awell-deï¬ned model will help to reduce the number of failed studies owing to design failure. Finally, if an unfavourable outcome is predicted from modelling of a 'registration trial', the programme may be terminated before investment into the next,more costly, phase of development.
Dose selection for Phase III pivotal trialsmay be based on the PKPD modelling and simulation analysis of the data from only a few studies [22-24]. Apart from the obvious use of efï¬cacy data (PD effect/biomarkers), safety data can also be incorporated into the model. Even if amodel is developed using data from fewer studies, it can still enable selection of optimal dose and sampling schedule
However, if modelling is used as a basis for dose selection fo registration trials, regulators should be comfortable with the selected dose and the rationale behind the plans for their use in conï¬rmatory trials. It is prudent to reach an agreement with the
regulators on dose selection for pivotal studies before their start This, in turn, highlights the need for a closer interaction between industry and regulators through all stages of drug development (dose selection based on modelling and simulation is now frequently discussed at the end of Phase II ameetingswith the FDA)
Beneï¬ts of modelling in late clinical development (Phase III)
Phase III studies provide the ï¬nal conï¬rmation of the efï¬cacy and safety of the tested drug in a wide patient population of interest. They provide the ultimate safety and efï¬cacy data for approval of drug's use in clinical practice.
Once the target effect is identiï¬ed, the focus of modelling and simulation can fully move on to the optimisation of the study design to demonstrate robustly the effect and reduce the risk of failed study design [25].Modelling and simulation can utilise both efï¬cacy and safety data to build adequate models. Modelling in Phase III can be used to
- Assess the impact of applicable covariates (patient subpopulations-demographics, co-morbidities, concomitant medication and so on).
- Validate the population PK/PD model.
- Establish or conï¬rm dose exposure-response relationship in target population/ subpopulation.
- Assess need for dose adjustment in special populations.
In the case where there are favourable efï¬cacy data provided from one large Phase III (pivotal) trial, the ultimate beneï¬t of modelling and simulation would be in obviating the need for a second large trial. Current expectations from most regulators are
for substantial evidence of effectiveness to come from at least two adequate and well-controlled (AWC) trials (Phase III). In Europe,the existing guidance (CPMP/EWP/2330/99) does not formally require two or more pivotal studies but suggests that, in most cases, a Phase III 'programme with several studies is the most
feasible way to provide the variety of data needed to conï¬rm the usefulness of a product in the intended population' [26].
The FDA Modernisation Act of 1997, and its subsequent relevant guidance, allowed for the use of exposure-response information in combinationwith a single pivotal clinical trial as sufï¬cient evidence of effectiveness [27,28]. This approach has, for example, been successfully used in registration of gabapentin for post-herpetic
neuralgia that was based on efï¬cacy data and exposure-response analysis and extrapolations fromthe two trials using different doses (i.e. efï¬cacydata for relevant doseswerenot replicated inthe second trial). For the ï¬rst time, exposure-response (PK/PD) analysis was used to establish a linkage across two clinical studies to provide conï¬rmatory evidence of dose response [29].
FDA requested that in order to replicate the clinical trial, this PK/PD analysis had to
withstand the same qualitative and quantitative review as the data froman AWC. The PK/PD results indeed conï¬rmed the evidence of efï¬cacyacross the studied doses sothat additional clinical trialswere not needed for approval. The Package Insert for gabapentin refers to 'PK/PD modelling that provided conï¬rmatory evidence of efï¬cacy across all doses'.
Adopting a new standard of a single clinical trial plus conï¬rmatory evidence would lead to more efï¬cient drug development. There still needs to bemore guidance on the standards required for 'conï¬rmatory evidence'. Recently, a case has been made that
convincing evidence of the pharmacologic mechanism of the clinical effect of a drug serves the same purpose as the second large clinical trial [30]; this evidence could come froma singlewell designed Phase II trial. The real cost saving beneï¬ts would be
achieved if this conï¬rmatory evidence was acceptable when generated from pooled data from earlier and smaller studies. In assessing the potential for pooling of the data factors, such as design of available studies, treatment duration and other relevant covariates, will need to be considered.
Finally, modelling and simulation could become the ultimate 'outcome' of Phase III development. Modelling and simulation provide the opportunity to simulate different trials using hundreds of patients' data in silico to predict trial outcomes. Some of the
simulated trial models have already been tested against outcomes of 'real' clinical trials, showing good accuracy with a very high correlation between the simulated and actual results [31,32]. Once a PK/PD model is developed, it could be tested prospectively in an AWC trial for its predictive value. If a model is sufï¬ciently pre-
dictive and accurate in this test, it could then be used to simulate the second large trial and its outcomes. Thus, the second large trial would be 'performed' only in silico. On the basis of adequate models, simulations could provide predictions for both efï¬cacy and safety data. The desired (and required) variety of data that customarily comes from several separate studies can be provided through a well-deï¬ned and validatedmodel. This approach would be easier to advocate for new compounds where a sufï¬cient amount of safety data have already been generated throughout
their clinical development and for new compounds where a large amount of safety and efï¬cacy data have been acquired on the basis of similar compounds (class) and used to validate themodel. Thus, reduction in number and/or size of Phase III studies could be seen as the ultimate potential beneï¬t of modelling and simulations in
drug development.
There are currently some efforts to build comprehensive models for different disease states (e.g. diabetes, cardiovascular disease, asthma) that use public domain medical knowledge in relevant disease to enable translation of short biomarker response into clinical outcomes [33,34]. Building of such diseasemodels requires
a large patients' database with a large variability in relevant covariates. Some of these data are already published in different epidemiological and outcome studies [31,32]. However, pooling patients' data from different databases that exist across industry would provide invaluable source of information for disease modelling. Industry may choose to form syndicates, similar to those advocated for validation and qualiï¬cation of biomarkers. The predictive value of such diseasemodels will be tested prospectively against large clinical outcome trials. If validated, they can greatly help Phase III design and target population selection across the industry.
Results
PKPD-Modelling from the competent authorities perspective (EMA, National competent authorities (EU), FDA)
PKPD-Modelling from the industries perspective (To what extent is PKPD-Modelling already used, Literature search, EPARs)
PKPD-Modelling in European Assessment Report
Bridion - EPAR - European Assessment Report EMEA/H/C/000885
Bridion (International Nonproprietary Name -INN - Sugammadex) is a novel compound that has been developed for reversal of neuromuscular block induced by the non-depolarizing neuromuscular blockers (NMB) rocuronium and vecuronium.
Bridion is a modified γ-cyclodextrin and its mode of action is based on the forming of 1:1 inclusion complexes with rocuronium or vecuronium. Upon intravenous injection complex formation reduces the amount of free neuromuscular blocking agent, leading to a fast reversal of neuromuscular block. It represents an entirely new approach in the management of reversal of neuromuscular blockade.
However, sugammadex does not reverse neuromuscular block induced by succinylcholine or benzylisoquinolium compounds.
This unique mechanism of action distinguishes Bridion from the class of anticholinesterase inhibitor reversal agents. Furthermore, selected NMBs (e.g., rocuronium) have been shown to have a higher affinity for Bridion than for the nicotinic receptor, allowing the reversal of a profound NMB to be possible. The mechanism of action of Bridion does not result in stimulation of the cholinergic nervous system. There is no need for concomitant administration of antimuscarinic drugs. Furthermore, due to the removal of the muscle relaxant from its site of action, Bridion is able to reverse even a very profound NMB rocuronium and vecuronium. [Z ]
The legal basis for this application refers to: Article 8.3 of Directive 2001/83/EC, as amended - complete and independent application.
In Summary 15 clinical trials have been conducted to evaluate the benefit of suggamadex in surgery and special populations. Reading the EPAR of Bridion it can be found, that PK-Modelling was used to gain more precise information on this novel compound regarding four subgroups in special populations.
Paediatric
Elderly (>65years, but <80 years)
Renally impaired
As suggamedex is excreted renally
Nplate - EPAR - European Assessment Report EMEA/H/C/942
Immune thrombocytopenic purpura (ITP) is an autoimmune disorder characterized by platelet destruction caused by antiplatelet auto-antibodies. The estimated incidence is 100 cases per 1 million people per year, and about half of these cases occur in children. 72 percent of patients over 10 years old are women, and 70 percent of these women are less than 40 years old. ITP is classified as acute (of six month or less in duration) or chronic. In addition it can be primary or secondary to an underlying disorder.
The diagnosis of ITP can be difficult because of the variety of presentations and lack of specific criteria. The only consistent abnormalities are microangiopathic haemolytic anemia, characterized by red-cell fragmentation, and thrombocytopenia, features that can also occur in other conditions. Neurologic and renal abnormalities and fever can also occur. The onset of the disease is often insidious because history and physical examination are normal except for symptoms that accompany platelet disorders (purpura, epistaxis, and gingival bleeding are common). In general ITP is defined by a low platelet count, normal bone marrow, and the absence of other causes of thrombocytopenia. Currently available treatments for chronic ITP act by decreasing platelet destruction including corticosterosteroids, intravenous gamma globulin (IVIG), intravenous ant-Rho(D) or suppressing the production of antiplatelet antibodies. Relapse is common when these agents are discontinued. Surgical splenectomy is generally considered second-line therapy for adults with ITP after the use of steroids and usually results in a remission. A small percentage of patients is severely affected and does not respond to treatment. Patients who fail to maintain a platelet count above 30 x 109 /L after 6 to 12 months of standard therapy are often defined asrefractory to therapy.
PKPD modelling and simulation will lead to fewer failed compounds, fewer study failures and smaller numbers of studies needed for registration, when used on the base of knowledge and establishing of respective guideline in EU, US and Japan - hence the ICH area. Intotal, there is more training necessary from agencies and
