Download PDF

Europace (2014) 16, 724–735
doi:10.1093/europace/euu009
FOCUSED ISSUE: REVIEW
Computer-based prediction of the drug
proarrhythmic effect: problems, issues,
known and suspected challenges
Barbara Wiśniowska 1*, Aleksander Mendyk 2, Kamil Fijorek 3, and Sebastian Polak 1
1
Unit of Pharmacoepidemiology and Pharmacoeconomics, Faculty of Pharmacy, Medical College, Jagiellonian University, Medyczna 9 Street, 30-688 Kraków, Poland; 2Department of
Pharmaceutical Technology and Biopharmaceutics, Faculty of Pharmacy, Medical College, Jagiellonian University, Medyczna 9 Street, 30-688 Kraków, Poland; and 3Department of
Statistics, Faculty of Management, University of Economics, Rakowicka 27 Street, 31-510 Krakow, Poland
Received 19 June 2013; accepted after revision 8 January 2014
It is likely that computer modelling and simulations will become an element of comprehensive cardiac safety testing. Their role would be primarily
the integration and the interpretation of previously gathered data. There are still unanswered questions and issues which we list and describe
below. They include sources of data used for the development of the models as well as data utilized as input information, which can come
from the in vitro studies and the quantitative structure–activity relationship models. The pharmacokinetics of the drugs in question play a
crucial role as their active concentration should be considered, yet the question remains where is the right place to assess it. The pharmacodynamic angle includes complications coming from multiple drugs (i.e. active metabolites) acting in parallel as well as the type of interaction with
(potentially) multiple affected channels. Once established, the model and the methodology of its use should be further validated, optimistically
against individual data reported at the clinical level as the physiological, anatomical, and genetic parameters play a crucial role in the drug-triggered
arrhythmia induction. All the abovementioned issues should be at least considered and—hopefully—resolved, to properly utilize the mathematical models for a cardiac safety assessment.
----------------------------------------------------------------------------------------------------------------------------------------------------------Keywords
Modelling † Simulation † hERG † Proarrhythmic potency † Cardiac models
Introduction
The primary goal in the drug development process is to provide to
the market safe and effective drugs meeting the needs of the patients.
However, bringing a new drug to the market, from discovery to approval, is a very long, complex, and costly process (Figure 1). High attrition rates, especially in the late phases, contribute substantially to
the total cost of developing a new drug. As non-clinical and clinical
safety issues are the leading causes of compounds termination and
withdrawals of marketed drugs, safety is a principal concern from
the earliest stages of the research and development process.
The recent literature reviews reveal that cardiovascular toxicity is
the most common reason for drug attrition in all the phases of development and clinical testing, and during post-marketing surveillance.1,2 Drugs and other xenobiotics can affect the heart tissue in
many different ways, i.e. they can cause direct myocyte injury, activate
apoptotic and necrotic changes, alternate ion homoeostasis or signalling pathways, or influence transcription factors (i.e. kinase inhibitors3,4). However, it is the prolongation of a QT interval on an
electrocardiogram (ECG) underlying life-threatening cardiac
arrhythmias—torsade de pointes (TdP), and the following ventricular fibrillation that is the most prominent cardiovascular adverse
effect. Proarrhythmic effects identified during post-approval use
were responsible for the withdrawal of several blockbuster drugs
in recent years and extensive ‘black-box’ warnings for many others.
In most cases, TdP is self-limiting and only occasionally may degenerate into a potentially fatal ventricular fibrillation. Nevertheless,
while most of the drugs capable of inducing TdP are used for the
symptomatic treatment of rather benign conditions (with the exception of antiarrhythmics), the associated risk of sudden cardiac death is
unjustifiable. Therefore, cardiac liability requires a careful scrutiny
during the drug development process and constitutes a significant
concern for the regulatory agencies, including the European Medicines Agency and the Food and Drug Administration. There is,
however, a growing awareness of its insufficiency and the need
for the evaluation of drug influence also on cardiomyocyte metabolism, structure, and function, including contractility or non-human
Ether-à-go-go-Related Gene (hERG) arrhythmia.
* Corresponding author. Tel: +48 12 620 55 17; Fax: +48 12 620 55 19, E-mail [email protected]
Published on behalf of the European Society of Cardiology. All rights reserved. & The Author 2014. For permissions please email: [email protected].
725
Challenges with computer-based prediction of the drug proarrhythmic effect
Compound discovery and screening
5 000–10 000
Compounds
2–10 years
Discovery
Laboratory and animal experiments
250
Compounds
3–6 years
• Target identification and validation
• HIT generation
• Lead development and optimization
• ADME, pharmacokinetics
• Parent compound assay
Preclinical
• In vitro toxicity
• Pharmacodynamics
• Assay metabolites
• In vitro toxicity
Phase II
Phase I
5
Compounds
6–7 years
Clinical
1 DRUG
Continues
throughout
drug life
Approval process
• FDA/EMA/MHLW evaluation
• Approval
Launch
1 DRUG
Continues
throughout
drug life
Phase III
• First in human
• Small scale • Large scale
studies
patient
patient
studies
studies
• Pharmacokinetics
• Efficacy, side
• Pharmacodynamics • Dosage
optimisation effects
• Safety
Post-marketing surveillance
Phase IV
• Additional post-marketing testing
• Pharmacovigilance
• Pharmacoeconomics
Figure 1 Stages of the drug discovery, development, and approval process. Source: based on the PhRMA Profile Pharmaceutical Industry 2013.
Safety pharmacology meets the modelling and simulations approach in many areas. A multitude of data coming from various
types of studies (in vitro and in vivo animals, in vivo humans) reveals a
clear need and an open arena for implementing meta-models (yet
another tool, in fact), helping in their interpretation and potentially
in the decision-making process. An example is the DILI-sim initiative
offering a set of in silico-realized computational models to predict
drug-induced liver injury in exposed patients.5,6 For the proarrhythmic potency prediction of drugs, biophysically based computational
mechanistic models of action potential (AP) formation have been
used for a long time. An analysis of the available reports and publications depicts mainly academic groups developing and testing various
solutions.7,8 More recently, however, there has been a growing interest of the industrial labs in the modelling approaches.8 – 12
Since the modelling and simulation approach does not aim to
replace the in vivo experimental models for cardiac safety testing, it
offers possibilities which are either difficult or even impossible to
gain at the pre-clinical and clinical levels of drug development due
to practical, meritorious, or ethical reasons. The capacity of virtual
testing to provide data that cannot be obtained experimentally is
726
probably one of the most recognized benefits of using the computational models of cardiac physiology at various levels of complexity.
The questions which can be addressed cover various areas, which
might be divided into several groups:
(1) Simulations at the population level
(a) inter-individual variability and its influence on the ECG traces
and their derivatives (QT/QTc/QRS)
(b) influence of the pathological changes on simulated ECG
traces
(2) Influence of the drug pharmacokinetics (PK) on the cardiac effect
(a) the effect of the projected human PK profile on the therapeutic index
(b) effect of the peak-to-trough ratio
(3) Influence of the drug PK and the pharmacodynamic (PD) interactions at various levels
(a) drug –drug and/or food and/or disease interaction
(4) Drug-triggered physiology disruption
(a) heart rate changes
(b) plasma ions concentration modification
(5) Other
(a) additional indirect drug effects on the currents (including the
up- or the down-regulation of channel trafficking/expression, and metabolic modulation)
(b) electrophysiological responses on cellular, tissue, and organ
levels given appropriate data, and thus increasing confidence
in risk assessment (and probability of success) of novel drug
therapies.
In the current review, we do not aim to compare the models of
human cardiac physiology either from the structure or the computational efficiency point of view. Readers interested in this topic are
encouraged to read the comprehensive review published relatively
recently by Clayton et al.13
Current approaches used for drug
proarrhythmic potency assessment
The proarrhythmic property of the compound can be assessed at all
the stages of drug development. In the 1990s, the cardiac safety of
non-cardiac drugs became a major safety issue for the industry, and
three guidelines regarding clinical (ICH E14—ICH guidance for the
industry on the clinical evaluation of QT/QTc interval prolongation
and proarrhythmic potential for non-antiarrhythmic drugs) and nonclinical (S7A—ICH guidance for the industry on the safety pharmacology studies for human pharmaceuticals; S7B— ICH guidance for
the industry on the non-clinical evaluation of the potential for
delayed ventricular repolarization (QT interval prolongation) by
human pharmaceuticals) testing strategies were developed and
adopted by the regulatory agencies. The non-clinical testing strategy
includes an in vitro IKr current assessment in the cell lines expressing
the hERG channels, the AP recordings in isolated myocytes and
tissues, as well as the in vivo QT assays measuring the various
indices of cardiac repolarization. At the early discovery and lead
B. Wiśniowska et al.
optimization phases, the hERG channel plays a pivotal role. It is also
important for the initial assay to be amenable to testing a large
number of compounds in reasonable time. Thus, at this stage,
methods offering higher throughput, even if non-good laboratory
practice (GLP) compliant, are preferred to the gold-standard
manual patch clamp technique (PC). Rubidium efflux, fluorescence,
and radioligand binding measurements are acceptable techniques,
in terms of the throughput and the false negatives, eligible for
primary screening, although some concerns have been raised about
their predictiveness.14 – 22 The GLP-compliant conventional PC
method enables the direct scrutiny of the timecourse changes of
the potassium channel current with a simultaneous control of the
voltage conditions and provides detailed, high-quality data. Nonetheless, it is labour intensive, requires special equipment, a skilled operator, and is relatively costly. The use of traditional patch clamping is
almost exclusively restricted to the late stages of the hit-to-lead
process and safety testing. However, the recent advances in automated PC technology bring the potential to provide a high information content at a sufficient throughput earlier in the discovery
process.23,24
Another pre-clinical test recommended by the ICH (International
Conference on Harmonisation of Technical Requirements for Registration of Pharmaceuticals for Human Use) guideline is the assessment of the AP duration (APD), potentially altered by the presence
of the tested compound. Both multicellular preparations and disaggregated cardiomyocytes can be used.
Single-cell models are suitable for assessing the effects on both
the ionic currents and the APD. High-quality data can be provided
by human ventricular cardiomyocytes, which offer a possibility of
hERG channel inhibition assessment within its natural environment,
with all the known or undiscovered subunits, ancillary units, and
other cellular factors modulating channel activity. However, there
is a lack of sufficient and viable biopsy material from humans, availability being usually limited to atria or diseased ventricle, thus despite
the interspecies differences in magnitude and sometimes the gating
kinetics of IKr, myocytes from other animal species, such as guinea
pig, dog, or rabbit, are used. The current advances of stem cell technology deliver a promise of new opportunities for the predictive cardiotoxicity assays going beyond the hERG and being amenable to
high-throughput assays. Stem cell technology could overcome the
problem with human material availability, and the disadvantages of
using animal cells preparations25 – 29; however, there are still issues
such as the heterogeneity of the stem cell-derived cardiomyocytes
and their foetal characteristics or the efficiency and the reproducibility of the differentiation protocols that need to be addressed before
stem cell technology could be adopted for drug discovery. Readers
interested in the details are referred to the excellent reviews published by Khan 201328 and Mordwinkin et al.30
The main advantage of the multicell preparation over the single-cell models is a better representation of the in vivo situation.
The best materials for the pre-clinical safety studies are probably
the primary human cardiomyocytes, which offer a possibility of the
hERG channel inhibition assessment within its natural environment.
Owing to the limited availability of human biopsy material, the myocytes retrieved from other species—including guinea pigs and dogs—
are used as a compromise. Still, the use of the myocytes is technically
demanding, not to mention the interspecies differences in magnitude
Challenges with computer-based prediction of the drug proarrhythmic effect
and sometimes the gating kinetics of IKr. For that reason, papillary
muscle or Purkinje fibre preparations are used as an alternative.
Yet, neither of them displays an identical composition of the ion
channel currents as compared with the human myocardium, which
results in a different drug sensitivity and the species-specific AP
shape and duration. Moreover, the higher the assay complexity, the
more difficult it is to interpret the results. Furthermore, even the
best animal model is unable to account for the inter-individual variability, while the strains and the species are selected to be maximally
homogeneous. Thus, caution should be exercised in extrapolating
the animal-to-human data. Nevertheless, the in vitro studies are generally faster, more flexible, and cost-effective than the in vivo animal
testing. They also gained the approval of the regulatory bodies due
to the ethical considerations of animal protection and welfare, and
scientific issues, such as a lack of concordance between the animal
models and the adverse human outcomes. Thus, the in vitro models
are extensively used, particularly to prioritize toxicity screening
and to predict the outcome of the later animal safety studies, and
finally eventual human-specific toxic effects. However, given that
the repolarization process is complex and is influenced by many parameters at multiple functional and anatomical levels, it is natural that
none of the in vitro and in vivo studies described briefly above are
solely sufficient for assessing the proarrhythmic potential and
making decisions, and these approaches should be considered complementary. Despite having inherent limitations, the pre-clinical evaluations offer predictive correlations with the in vivo outcomes and
provide an opportunity not only to identify and reduce or eliminate
a hazard before administration to humans, but also to reduce
animal use during experiments, thus they constitute an element of
the integrative cardiac safety assessments. The pre-clinical proarrhythmia models, with their strengths and limitations, are discussed
in detail by Raschi et al.,31 to whom the interested readers are
referred.
Regardless of the assays employed at the earlier phases of the preclinical studies to detect or predict cardiac liability, in vivo animal
studies are still needed before the clinical trials. Although providing
the best available data regarding human proarrhythmic risk for drug
dose selection in the first-in-human studies, the animal models are
still an imperfect predictor of the torsadogenic risk in humans. Nonrodent (dog, monkey, swine, rabbit, ferret, and guinea pig) QT assessment meets the objective of both ICH S7A (core battery study) and
S7B. In vivo intact animal models allow the investigation of drug proarrhythmic effects with the influence of metabolism, hormonal, and
nervous systems accounted for. The preferred methodology is the
use of unrestrained telemetered animals, usually dogs. The principal
endpoint in these studies is the QT interval length, but frequently also
the other safety parameters, such as the heart rate, blood pressure,
and ECG, which are measured simultaneously. The disadvantages
of the animal studies include, but are not limited to, the costs and a
low-throughput capability. The main scientific challenge remains
the results translation from the animal-to-human situation due to a
lack of some electrophysiological effects in various species.32 The
problem of the animal-to-human scaling becomes even more challenging when the inter-individual variability is considered.
A pre-approval clinical development programme provides a rigorous assessment of the drug’s propensity to prolong the QT interval. Thorough QT studies (TQTs) involve the quantification of
727
the degree of the drug’s influence on cardiac repolarization in
healthy volunteers as compared with placebo and a positive control.
The aim of the TQT studies is to identify drugs that need more attention for their cardiac liability and require an additional ECG monitoring
in subsequent trials to assess the arrhythmia risk in the target population. Although the TQT studies are informative and the best among
the currently available methods, they are costly and suffer from a
low positive-predictive value.33 – 35
Computer modelling and
simulations—components of the
comprehensive cardiac safety test
Cardiac simulation is no longer an interesting yet not very useful curiosity, and has its place in various stages of drug discovery, design, and
safety assessment.36 – 41 It is obvious that the modelling and simulation approach offers new possibilities in the area of proarrhythmic
potency assessment. There are still unanswered questions and
issues which we list and describe below. They should be at least considered and—hopefully—solved, to properly utilize the mathematical models for a cardiac safety assessment.
In vitro studies results as the source of input
The predictiveness of the mathematical models is highly dependent
on the quality of the input data. There are guidelines and recommendations rather than strict regulations that aid a pre-approval proarrhythmic potential assessment. Specific tests and their design are
not institutionally defined, and the choice of the specific elements
of a testing strategy belongs to the researcher.42,43 Thus, the
results from different laboratories are difficult to compare due to
the variations in methodology and also their incorporation into
the mathematical models is not straightforward. A wide variety of
pre-clinical in vitro studies can be employed to evaluate the drug–
hERG interactions potential. The agreed gold standard for the
hERG liability evaluation is the manual PC technique. Manual PC, although providing high-quality data, is not suitable for screening large
amounts of molecules. More convenient, especially for initial screening, are automated multiwall plate format methods, which offer a
high-throughput, time, and fund savings. These methods, if validated,
can offer a substantial potential for generating large and robust datasets for modelling. A recent study of Di Veroli et al.44 is an excellent
example of applying the high-throughput screening data in drug safety
assessment. Regardless of the technique type, the variability of the
measured half-maximal inhibitory concentration (IC50) values
results from many other experimental parameters. One of the
major sources of data discrepancy in the literature arises from the
use of different test systems. While most drug– hERG interactions
models are based on the experimental results from mammalian
cells (data retrieved with the use of the XO cells are not included
into the datasets), other differences, e.g. HEK vs. CHO or hERG 1a
vs. 1a/1b subunits expression and transfection technique are not considered. Another factor of possible importance in the compound
potency assessments is the temperature. The IC50 values retrieved
both at room and physiological temperatures can be found in the literature, however although the temperature dependency of the
hERG inhibition has been pointed out, the measurements at
728
B. Wiśniowska et al.
EXPERIMENTAL DESIGN
•
•
•
•
•
•
•
•
•
•
•
•
Assay type
Temperature
Voltage protocol
Stimulation frequency
Bath solution
composition
pH
Devices: automatic/
manual: type
User
Number of cells
analysed
Number of
concentrations for
concentration-inhibition
curve construction
Concentration range
Fitting method
DRUG
MODEL
• Complexity (e.g. cell,
whole myocyte,
tissue)
• Biological variability
• Expression system
• Subunits
• Transfection method
• Physico-chemical
properties
• Nature of block
Experimentaly measured
IC50 value
Figure 2 Variability sources of the in vitro-based hERG assays.
physiological temperature become more and more prevalent.43,45
The predictive value of the in vitro studies can be further affected by
many more factors, some of them presented in Figure 2, thus the
data from various laboratories are not concordant. In view of the
foregoing, it seems to be of the utmost importance to follow a
clearly defined methodology in the model establishment process;
either to not introduce noise or inconsistent data or to give a possible
reasoning for the eventual discrepancy between the simulation
results and the clinical observations.
Quantitative structure –activity relationship (QSAR) models correlating the chemical structure and the in vitro-measured ionic current
inhibition are tools which can be utilized in a situation where there are
no measured data available. However, all the concerns mentioned
above regarding the lack of a standard methodology can be applied
here as well. Since the information used for the development of
the models is very often highly heterogeneous and the modelling
methodology varies significantly between the studies (i.e. models validation stage), the calculated results can be biased.
In vitro—in vivo scaling: under what settings
are the in vitro models (cells)
representative of the human physiological
conditions?
The in vitro results constitute a vast and irreplaceable source of data
for the in silico cardiotoxicity modelling. However, in the majority
of the cases, the in vitro-measured results are directly transferred to
the in silico system without further transformation. Such an approach
seems to be biased considering that the variability of the in vitro results
is significant and depends strongly on the study settings. Not only
with regard to the in silico models performance but also for the
direct comparisons of the in vitro readouts and their predictive
value, the default settings should be defined. As a general rule, the
most appropriate conditions for the safety studies aimed at providing
the estimates of human risk, should replicate as close as possible the
human in vivo situation. Optimally, the native human cardiac ion channels in their natural environment should be used. On the other hand,
technically demanding, low-throughput, labour intensive, or costly
tests (even if they are most accurate) may be inconvenient or inapplicable to each stage of the drug development. The technique of first
choice must satisfy both the predictiveness level demanded at a
given stage and the resource consumption. As such, the investigations
of the drug-triggered current fluctuations involve cloned human ion
channels heterologously expressed in mammalian cells. Intuitively,
human embryonic kidney (HEK) cell lines should be preferred, due
to their human origin. The issue of the temperature dependence
of the hERG channel gating kinetics was recognized some time
ago,45 – 49 with the conclusion that it more closely resembles the
endogenous IKr when studied at a physiological temperature.49,50 The
era of automation in electrophysiological experimentation has brought
new technical possibilities; as a consequence, it is now common to
perform assays at a physiological temperature. Furthermore, to
improve the relevance of the in vitro data and provide a more accurate
cardiac risk assessment, potassium (and other relevant) ion concentration in the bath solution ought to correspond with the ranges occurring
in the physiological conditions; holding potential and depolarization
voltage must meet the voltage ranges of the physiological AP, and the
stimulation frequency should be tuned so as to represent in the best
manner possible the human in vivo situation. Quinn et al.51 have recently
published a draft Minimum Information about a Cardiac Electrophysiology Experiment, an initiative with the ultimate goal of developing the
standards for recording, annotating, and reporting experimental data,
which is meant to facilitate the utilization of findings between labs and
data incorporation into the models.
Before the development and the implementation of these standards, a possible solution attempting to deal with the in vitro
729
Challenges with computer-based prediction of the drug proarrhythmic effect
experimental data inconsistency is to apply a kind of scaling factors
that would serve to extrapolate the measured results to the agreed
standard conditions, and then use the modified dataset in the modelling process.51 The task is not trivial, because the relationship can be
non-linear, and the availability of proper data is limited. However,
some efforts have been made to establish the scaling factors for
some of the results variability sources.52,53 Such a set of extrapolation
factors can allow for the exploitation of the inconsistencies in the
model development process (while considering some of the elements of the experimental procedure settings) and the experimental
data without a depreciation of the predictions of the model.
Source of the data used for the
mathematical models development
The main pragmatic aim of using the mathematical models is to translate the pre-clinical data into the clinical situation in humans without
the need for the animal models utilization and thus avoiding a potentially problematic allometric scaling. Therefore, to be able to extrapolate the in vitro studies findings to the human situation, an
accurate model for the human ventricular AP, which reproduces a
broad range of physiological behaviours, is needed. There is a plethora of available models describing the cells electric physiology of
various heart parts (Purkinje fibres, atria, and ventricles). The majority of the models are based on mosaic data which come from various
sources and donors, and the differences between the cell types and
the species affect the results. It is a well-known and commonly utilized
approach used since the McAllister–Noble– Tsien model of the Purkinje fibre was published in 1975.54 However, there should be a clear
recognition of the potential bias of the simulation results being a consequence of the inhomogeneous in vitro study settings used at the
model development stage. Cherry and Fenton compared the mathematical models of canine and human ventricular myocytes, and
human atrial myocytes, and give some possible explanations for a disagreement among the predictions of the models of the same species
and region of the heart. The implications of the differences between
the models for their applicability in the computational studies of electrophysiology were also discussed.55
Cellular electrophysiology experiments, crucial for the understanding and a further mathematical description of the cardiac physiological mechanisms, are usually performed with the channels
expressed in human non-myocytes (i.e. HEK cells). The models,
listed among the major achievements and milestones in the development of the electrophysiology of the mammalian cardiac ventricular
models, have been widely used until today, including a cardiac safety
assessment, although a question remains as to whether the results
reflect human physiology and can be directly used to draw conclusions and make binding decisions.56 – 58 Niederer et al.59 presented
a meta-analysis of two cardiac electrophysiology models investigating
the origin of the experimentally derived parameters. The phylogenetic trees of the ten Tusscher and Iyer models imply the diversity of
the data sources used for the model derivation.41,60,61 A subsequent
sensitivity analysis compared the functional significance of the
sodium-potassium pump for defining the restitution curves. The
obtained results indicate that even though the models aim to represent the same physiological system, the functions of the equivalent
components are significantly different. O’Hara and Rudy46 in their
recent study quantitatively assessed the response to drugs in
human and non-human species with the use of the computational
models of human, dog, guinea pig, and rabbit ventricular epicardial
cells. It was concluded that different mammalian species react differently to the drug-triggered cell electrophysiology modifications: currents reduction (IKr and IKs) and b-adrenergic stimulation. What is
even more important is that the early after-depolarization formation
due to a delayed rectifier K+ current block at slow pacing rates is
species dependent. After considering this, the authors suggest
caution when extrapolating the results of the drug effects tested on
the non-human species to the safety and efficacy in human clinical application. One of the abovementioned aspects, namely b-adrenergic
stimulation can be addressed by coupling the models of the electrophysiological and b-adrenergic elements of the heart.62
Physiological parameters—simulation
at the individual level
There are considerable concerns about the predictive value of the
pre-clinical TdP markers. While TdP arrhythmia is always preceded
by QT interval prolongation, not every case of QT prolongation is
harmful. The abovementioned and the extremely rare occurrence
of TdP suggest that the human physiology-dependent parameters
will play a pivotal role in the clinical manifestation of the acquired
long QT syndrome. The estimation of the inter-individual variability
in the clinical manifestation of the drug-induced electrophysiological
changes during the pre-clinical studies represents a big challenge. To
date, the majority of the current mathematical models of the cardiac
cells or the heart tissues employ fixed physiological parameters
values to model an outcome. In general, the mean values are used
to predict average patient response (average QT interval prolongation) and the approaches aimed at assessing the outcomes distribution in the population are rather limited, however a predictive
model should combine the pre-clinical in vitro data with the information on the demographic, physiological, anatomical, and genetic variability to allow a reliable extrapolation to the in vivo situation. More
recently, a growing number of studies have aimed at investigating
the potential sources of inter-subject variability and the following differences in the electrophysiology of the heart and account for them in
a mathematical modelling of the cardiac effects at the population
level.63 – 65 The authors of this review proposed a QT prolongation
model accounting for the age and the sex-related differences of the
cardiomyocyte structure and the main serum ions concentrations;66,67 however, these are only a few of the many parameters
which may affect the outcome expression at an individual level. The
actual risk value for a particular patient is also not constant and fluctuates according to individual circadian rhythms (heart rate, ions
plasma concentration), or rhythms with a longer periodicity (sex hormones). The models of the physiological, genetic, and anatomical
parameters variations would allow the generation of a virtual
human population which could mimic the real clinical situation.
A real challenge is to simulate a drug effect for the target population
that would not be fully represented by the rigorously selected clinical
trial samples, strictly following the study protocols. During the postmarketing phase, when the uncovered TdP cases are reported
(during the pre-approval process), the prescription and use practices
are far from the standardized trial protocols; a drug is used by
730
diseased individuals with co-morbidities, treated with multiple drugs,
and often not fully adherent; the target population can incorporate
elderly patients with pathological changes and age-dependent physiology modifications and special populations such as pregnant women
or children. One of the prominent examples of model parameterization to study the potential sources of variability in the electrophysiological measurements is given in a recent publication by Walmsley
et al. The authors used the mRNA expression data to predict the
functional differences in the AP biomarkers between a failing and a
non-failing human heart.65 A reader might be also interested in the
review by Nattel et al.68 dealing with the mRNA profiling and
the role of such data in an assessment and an understanding of the
cardiac electrical function modification in both physiological and
pathological situations. While it is unfeasible to test all the possible
scenarios during the clinical trials, the in silico simulations would
allow not only the improvement of the planning stage and the
running human trials, but would also enhance the predictive value
of a clinical trial and authenticate the results extrapolations for a reallife target population.
What effect does genetics have?
An assessment of the genetic constitution can be crucial for a proper
drug safety assessment at the population level.69 The majority of
serious or fatal cases of drug-triggered arrhythmias (or in an even
broader sense—cardiotoxicity cases) appeared after a relatively
long drug presence on the market.70 It indirectly suggests that such
a rare effect can have an idiosyncratic character and be connected
with the genetically determined parameters rather than the doseand the concentration-dependent adverse effects. As in the PK,71 a
significant clinical influence of the genetic constitution can be also
expected at the pharmacodynamic –toxicodynamic angle, where
either the functionality or the density of the drug target, modified
by the genetic constitution, defines the drug activity and safety. Considering that significant genetic modifications occur rarely in the
population, an anticipation of their clinical effects from the clinical
trials results is unlikely, therefore in silico modelling and simulation
offers complementarity to the clinical observation tool allowing for
the assessment of the functional consequences of the gene mutations
at various levels from cell to organ with the use of various algorithms
including classical Hodgkin–Huxley paradigm and Markov chain
models.72
It has been suggested that single nucleotide polymorphisms and, to
an even greater degree, mutations in the hERG channel can be associated with the QT interval abnormalities triggered by non-cardiac
drugs. The frequency of occurrence of the chosen polymorphic
forms of the KCNH2 gene (hERG) encoding the a-subunit of the
Kv11.1 potassium ion channel can be significant, reaching up to
30% of the heterozygous frequency in individuals otherwise recognized as healthy.73 – 76 It should be taken under consideration
during a simulation run at the population level, despite the fact that
the pharmacogenomic studies on the relation between the
KCNH2 polymorphisms and the sensitivities to drugs were mostly
conducted in the in vitro settings and still need a clear clinical
proof.77 There have been several attempts and mathematical apparatus involved in the mimicking of the various polymorphic forms of
the ionic channels.78 The elements of the genetic variability are the
isoforms of the ionic channels. For hERG1, two isoforms have been
B. Wiśniowska et al.
detected, namely hERG1a and hERG1b with the latter one
(hERG1b) found in 10 –20% of healthy individuals, based on the
mRNA quantification. Most importantly, ERG1a and ERG1b homomers, as well as the ERG1a/b heteromer, differ in their kinetic properties, which could be further connected with the potential long QT
syndrome.46 Therefore, a mathematical description and the potential
consequences on the simulation outputs would be desirable to properly mimic the human situation.79 The hERG channel is obviously not
the only one whose genetic modification can influence the clinically
observed ECG and its time derivatives. Among others, the mutations
in the KCNQ1 of the IKs channel and SCN5A of the hNav1.5 channel
can be simulated and their influence on the cardiac effects of drugs
quantitatively assessed.80 – 82
Multiple ion channels affected
The hERG inhibition, being responsible for the vast majority of the
drug-related excessive QT interval prolongation and/or the TdP arrhythmia cases, is the earliest indicator of the torsadogenic risk in the
compound development process. Yet, it is important to emphasize
that the studies limited to only the hERG channel may not fully
reflect the drug potential in the modifying ion currents. The latter
is due to the possible compound concomitant influence on the
other ionic channels of the cardiomyocyte, which can result in
either a potentiation or a mitigation of the clinical effects of the
hERG block. Potent hERG channel antagonists, such as verapamil,
amiodarone, propafenone, or flunarizine,8,83,84 with no QT prolongation effect on the one hand, and weak inhibitors or drugs
devoid of the hERG liability, such as alfuzosin, sotalol,85 associated
with the QT prolongation and TdP on the other, prove that the clinical effects of the drugs influencing heart activity is a complex process.
The effects of the hERG block are mitigated as a consequence of a
calcium channels block or by the simultaneous inhibition of the
inward late sodium current.83,86 The opposite effect may be the
result of an amplification of the outward currents during AP repolarization, with the prominent role of the slowly activating delayed rectifier potassium current (IKs), which is sometimes recognized as an
overlooked target in drug safety screening.87 The above gives some
reasoning for the lack of a straight correlation between the hERG inhibition and the QT interval prolongation. It is now widely recognized
that a block of one of the potassium channels is important, but
not sufficient, to predict the torsadogenic risk, thus the need for
the consideration of multichannel interactions of the compounds
arises.8,88,89 Such a procedure is repeatedly recommended, yet not
required.87 Novel high-throughput and automated techniques
using cells lines make it feasible to screen multiple ion channels
during drug development, thus there is a need to integrate these
data. The in silico models of cell electrophysiology, individual or
coupled into tissue or even whole-heart models, can be used to investigate the effects of the drug–multiple channels interactions on
the ionic currents, AP, and surface body ECG. In a recently published
paper, Kramer et al.,90 had confirmed a hypothesis that the empirical
models based on multiple ion channel inhibition assessment (hERG,
Nav1.5, and Cav1.2 currents) result in a better discrimination
between safe and unsafe drugs providing more accurate predictions
on the proarrhythmic potential. In the approach proposed by
Kramer, the profiles of the hERG, Cav1.2, and Nav1.5 ETPC indices
(defined as channel IC50/ETPC, where ETPC is the effective free
731
Challenges with computer-based prediction of the drug proarrhythmic effect
therapeutic plasma concentration) are used to distinguish between
the torsadogenic and the non-torsadogenic compounds. Such multiple ion channel effects based approach can be further utilized for
the computer-based simulation of the drugs influence on the ECG
and its time derivatives as well as a torsadogenic risk prediction.8,91
There is although a clear need of the standardized in vitro measurement procedure to minimize the inter-lab variability. Furthermore,
there are also other mechanisms and phenomena that can play a
role in the drug–ion channel interactions (protein expression and
trafficking, channel agonists), and influence the clinical appearance
of the proarrhythmic potency, which while considered, probably
could provide an even more reliable risk estimation and could
explain the weak correspondence of the extent of the QT interval
prolongation and the clinical manifestation of the TdP arrhythmia.
However, the relevant data are very sparse and much effort is
needed to develop models for alternative drug–ion channel interaction paths and integrate them with the existing electrophysiological
models.
How important is it to properly model
the drug –channel interactions?
This is one of the main elements influencing the quality of the simulation results, which needs more attention especially for drugs
which can interact in a complex way with a channel. Various types
of mathematically described interactions not only include simple
pore block on the one hand but also include more sophisticated
models based either on the guarded (GR) or modulated (MR) receptor theory (Figure 3).
The simple pore block is useful under the assumption that the
channel gating processes and the drugs do not interfere with each
other and the drug blocking activity does not depend on the
channel state. In this case, the drug binds to and blocks the channel
regardless of its state. An effect is simulated as a decrease in
maximal conductance, by the use of the scaling factor, which
depends on the drug concentration, and is described by the in vitromeasured IC50 value.
Both the GR and the MR receptor theories consider the various
states of the channels of interest (rest, inactive, active) and their
dynamic transitions described by the transition rates. The models
assume that a drug can bind to any channel state and the channel
with the drug bound is non-conducting until the drug unbinds from
the receptor.92 – 95 The GR and the MR hypotheses are mathematically identical; they differ only in the causes for state dependency of the
drug-binding process. In the MR model, access to the channel remains
constant but the affinity of a drug-binding region changes with the
channel state. In the GR model, the affinity of a drug-binding area is
constant, while access to the receptor varies with the channel state.
In practice, a simple pore block model is probably the most commonly utilized approach due to its simplicity.7 In addition, the chosen
drugs can act differently than by simple inhibition (i.e. induction,
certain state prolongation, e.g. open state for the Na channels).
What is the effect of multiple drugs
on the ion channels?
Multiple drugs effect on the electrophysiology of the cardiac myocytes is probably one of the most commonly met clinical situations,
and it is often neglected. The active metabolites at the clinical
studies level and polypharmacy at the clinical/ambulatory level are
the examples of the situations where such a scenario has to be considered. Here, we discuss a situation where the interaction results
from the same binding targets for multiple substances and subsequent currents modification.96 Simple sum of action is most
commonly used but it would be desirable to test it at the in vitro
level to assess the interaction mechanisms as the effect can differ
substantially from the abovementioned assumption as presented
by Friemel and Zunkler.97 The ultimate aim of all the drug safety
studies is to assess whether and to what degree some safety concerns
can be expected. As in the majority of the cases, the cardiac effects are
concentration-dependent and the idiosyncratic reactions (which
tend to be dose- and concentration-independent) occur rarely, the
range of the tested drug concentrations that should be considered
has to be precisely defined. There are, however, further elements
which are often overlooked and may be the source of erroneous predictions. It includes (potentially) the active metabolites of the parent
compound or other concomitantly taken drugs, and their interactions with the ion channels.
The role of the drug pharmacokinetics
A question remains about where the active concentration should be
assessed (measured). Plasma is the most convenient and relatively
easily available surrogate, although for some of the compounds,
active concentration at the site of action would be more appropriate
for modelling the drug effects on the cell electrophysiology, and in the
case of the ionic channels either heart tissue or pericardial fluid could
be considered.98 It has been reported that depending on the drugs
character, the serum to pericardial fluid concentration can be
significantly different than 1.99,100 As a direct measurement of the
drug concentration in a medium different than plasma would be
challenging if at all possible, a simulation of the active concentration
seems to be a potential solution. Proof-of-concept work showing
the ability of the in vitro—in vivo extrapolation approach, combining
the PK and the PD simulations of the clinical outcomes, has already
been presented.91 In the cited publication, a plasma concentration
was simulated and used as a surrogate. However, it is possible to
simulate the other tissues with the use of the physiologically based
pharmacokinetics approach.
Lack of appropriate clinical data
for the simulation efforts validation
Another practical challenge lies in the ability to robustly verify the
applied models performance. Unfortunately, a lack of clinical
studies results reported at the individual level make the simulation
efforts validation challenging and potentially biased. To the best of
the authors’ knowledge, all the previously reported comparisons of
the simulated and clinically observed drug-triggered effects were
based on the average reported values of the drug concentration
and the clinical effect. Since inter-individual variability resulting
from the differences in demography, physiology, and genetics
between individuals can influence the results significantly, it would
be desirable to precisely mimic the population involved in the clinical
study.
732
B. Wiśniowska et al.
Modulated receptor
V1
Simple pore block
[O]
[C]
[I]
Dk1
[R]
[DR]
Dk2
Dk1
Dk2
[DC]
Dk3
Dk4
[DO]
Dk5
Dk6
[DI]
V2
k1–rate
constant of drug association
constant of drug dissociation
D–drug concentration
[R]–receptors concentration
[DR]–bound ligand-receptor
complexes concentration [DI]-bound
ligand-inactivated receptor complexes
concentration
V1,2–constant describing voltage
dependent kinetics of channel gating
k2–rate
k1,3,5–state
dependent rate constant of drug association
dependent rate constant of drug dissociation
D–drug concentration
[C]–closed receptors concentration
[O]–open receptors concentration
[I]–inactivated receptors concentration
[DC]–bound ligand-closed receptor complexes
concentration
[DO]–bound ligand-opened receptor complexes
concentration
[DI]–bound ligand-inactivated receptor complexes
concentration
k2,4,6–state
Figure 3 Simplified representation of a simple pore block (left panel) and modulated receptor hypothesis (right panel). Simple pore block—drug
has permanent access to the target and the affinity of the drug receptor to the target is state independent; modulated receptor—drug (D) can bind to
the channels in the open (O), inactivated (I), or closed (C) state with different, state-specific affinities. Kinetics of the channel gating for the drug-bind
and the drug-free channels is voltage dependent (V1, V2), thus the dissociation rates (Dkn) are also voltage dependent.
Simulated endpoints
Conclusions
There are several problems connected with this element and one
of them plays a crucial role in defining the usability of the in silico
methods. The desired endpoint is a risk of arrhythmia (in most
cases—the TdP-type ventricular arrhythmia degenerating to ventricular fibrillation) and the simulated outputs include ECG and its
modification. It is a well-known fact that TdP is NOT a yes/no
effect and the QT prolongation is just a surrogate (probably not
the best one) of the TdP risk. Therefore, additional endpoints
would be desirable to assess the arrhythmia risk better. One of the
possible options is an electromechanical window recently described
as a promising surrogate and tested on the animal models.101,102 It
would require the implementation of the models coupling electric
and mechanical heart activity and expanding the current space of
the simulated endpoints.103
Another question that still remains open is the simulated signal
processing and its derivatives analysis. The QT/QRS calculation
methods can significantly influence the results, and the differences
can bias the conclusions drawn from the obtained results. Depending on the ECG analysis method applied, the results can
differ; therefore, standardization or at least a clear description
of the differences between the available methods would be also
required.
Several reviews have discussed the place and the potential for
modelling and simulation in the drug discovery process and safety assessment.8,36 – 40,104 – 108 The current review focuses on the problems connected with the implementation and the utilization of the
computational cardiac electrophysiology approach for a drug safety
assessment, nevertheless it still aims to show the potential lying in
the mathematical models and the modelling and simulation approach.
The abovementioned problems were divided into several groups and
include the technical and the biological obstacles, which one has to
challenge in real life. There is, however, another element which has
not been discussed and is probably as important as all the abovementioned—the problem of perception. The biophysically detailed
models of the cardiac physiology are sometimes incorrectly associated with the QSAR models as they are both realized in silico and
probably the in silico umbrella is a source of the confusion. In the
proarrhythmic potency assessment, the endpoint in the majority of
the cases is defined as the in vitro-measured IKr current inhibition, although there are some examples where the TdP risk is directly calculated.109,110 This empirical approach can be very useful especially at
the early stage of the development, yet cannot be compared with
the mechanistic, biophysiologically robust models describing the
human cardiac cells physiology. The main questions which should
Challenges with computer-based prediction of the drug proarrhythmic effect
probably be asked as a consequence are: How should the models be
used? What questions could be addressed? What input data are
needed to expect useful results?
Conflict of interest: S.P. is an employee of Simcyp Ltd. (part of
Certara) which is developing a platform for the drugs cardiac safety
assessment (CSS—Cardiac Safety Simulator).
References
1. Redfern WS, Ewart L, Hammond TG, Bialecki R, Kinter L, Lindgren S et al. Impact
and frequency of different toxicities throughout the pharmaceutical life cycle.
Toxicologist 2010;114(S1):1081.
2. Peters MF, Scott CW, Ochalski R, Dragan YP. Evaluation of cellular impedance measures of cardiomyocyte cultures for drug screening applications. Assay Drug Dev
Technol 2012;10:525 –32.
3. Force T, Kolaja KL. Cardiotoxicity of kinase inhibitors: the prediction and translation of preclinical models to clinical outcomes. Nat Rev Drug Discov 2011;10:
111– 26.
4. Mellor HR, Bell AR, Valentin JP, Roberts RR. Cardiotoxicity associated with targeting kinase pathways in cancer. Toxicol Sci 2011;120:14 –32.
5. Bhattacharya S, Shoda LK, Zhang Q, Woods CG, Howell BA, Siler SQ et al. Modelling drug- and chemical-induced hepatotoxicity with systems biology approaches.
Front Physiol 2012;3:1– 18.
6. Woodhead JL, Howell BA, Yang Y, Harrill AH, Clewell HJ III, Andersen ME et al.
An analysis of N-acetylcysteine treatment for acetaminophen overdose using a
systems model of drug-induced liver injury. J Pharmacol Exp Ther 2012;342:529 –40.
7. Obiol-Pardo C, Gomis-Tena J, Sanz F, Saiz J, Pastor M. A multiscale simulation
system for the prediction of drug-induced cardiotoxicity. J Chem Inf Model 2011;
51:483 –92.
8. Mirams GR, Cui Y, Sher A, Fink M, Cooper J, Heath BM et al. Simulation of multiple
ion channel block provides improved early prediction of compounds’ clinical torsadogenic risk. Cardiovasc Res 2011;91:53 –61.
9. Beattie KA, Luscombe C, Williams G, Munoz-Muriedas J, Gavaghan DJ, Cui Y et al.
Evaluation of an in silico cardiac safety assay: using ion channel screening data to
predict QT interval changes in the rabbit ventricular wedge. J Pharmacol Toxicol
Methods 2013;68:88 –96.
10. Bottino D, Penland RC, Stamps A, Traebert M, Dumotier B, Georgiva A et al. Preclinical cardiac safety assessment of pharmaceutical compounds using an integrated
systems-based computer model of the heart. Prog Biophys Mol Biol 2006;90:414 –43.
11. Davies MR, Mistry HB, Hussein L, Pollard CE, Valentin JP et al. An in silico canine
cardiac midmyocardial action potential duration model as a tool for early drug
safety assessment. Am J Physiol Heart Circ Physiol 2012;302:H1466 –80.
12. Gavaghan CL, Arnby CH, Blomberg N, Strandlund G, Boyer S. Development, interpretation and temporal evaluation of a global QSAR of hERG electrophysiology
screening data. J Comput Aided Mol Des 2007;21:189 –206.
13. Clayton RH, Bernus O, Cherry EM, Dierckx H, Fenton FH, Mirabella L et al. Models
of cardiac tissue electrophysiology: progress, challenges and open questions. Prog
Biophys Mol Biol 2011;104:22 –48.
14. Fermini B, Fossa AA. The impact of drug-induced QT interval prolongation on drug
discovery and development. Nat Rev Drug Discov 2003;2:439 –47.
15. Netzer R, Ebneth A, Bischoff U, Pongs O. Screening lead compounds for QT interval prolongation. Drug Discov Today 2001;6:78 –84.
16. Netzer R, Bischoff U, Ebneth A. HTS techniques to investigate the potential effects
of compounds on cardiac ion channels at early-stages of drug discovery. Curr Opin
Drug Discov Dev 2003;6:462 – 9.
17. Scott CW, Wilkins DE, Trivedi S, Crankshaw DJ. A mediumthroughput functional
assay of KCNQ2 potassium channels using rubidium efflux and atomic absorption
spectrometry. Anal Biochem 2003;319:251 –7.
18. Yang T, Roden DM. Extracellular potassium modulation of drug block of IKr. Implications for torsade de pointes and reverse use-dependence. Circulation 1996;93:
407– 11.
19. Rezazadeh S, Hesketh JC, Fedida D. Rb+ flux through hERG channels affects the
potency of channel blocking drugs: correlation with data obtained using a highthroughput Rb+ efflux assay. J Biomol Screen 2004;9:588 –97.
20. Tang W, Kang J, Wu X, Rampe D, Wang L, Shen H et al. Development and evaluation of high throughput functional assay methods for HERG potassium channel.
J Biomol Screen 2001;6:325–31.
21. Dorn A, Hermann F, Ebneth A, Bothmann H, Trube G, Christensen K et al. Evaluation of a high-throughput fluorescence assay method for hERG potassium channel
inhibition. J Biomol Screen 2005;10:339 –47.
733
22. González JE, Oades K, Leychkis Y, Harootunian A, Negulescu PA. Cell-based assays
and instrumentation for screening ion-channel targets. Drug Discov Ther 1999;4:
431 –9.
23. Stoelzle S, Obergrussberger A, Brüggemann A, Haarmann C, George M,
Kettenhofen R et al. State-of-the-art automated patch clamp devices: heat activation, action potentials, and high throughput in ion channel screening. Front Pharmacol 2011;2:76.
24. Yajuan X, Xin L, Zhiyuan L. A comparison of the performance and application differences between manual and automated patch-clamp techniques. Curr Chem Genomics 2012;6:87 –92.
25. Braam SR, Tertoolen L, van de Stolpe A, Meyer T, Passier R, Mummery CL. Prediction of drug-induced cardiotoxicity using human embryonic stem cell-derived cardiomyocytes. Stem Cell Res 2010;4:107 – 16.
26. Caspi O, Itzhaki I, Arbel G, Kehat I, Gepstien A, Huber I et al. In vitro electrophysiological drug testing using human embryonic stem cell derived cardiomyocytes. Stem
Cells Dev 2009;18:161 –72.
27. Dick E, Rajamohan D, Ronksley J, Denning C. Evaluating the utility of cardiomyocytes from human pluripotent stem cells for drug screening. Biochem Soc Trans
2010;38:1037 –45.
28. Khan JM, Lyon AR, Harding SE. The case for induced pluripotent stem cell-derived
cardiomyocytes in pharmacological screening. Br J Pharmacol 2013;169:304 –17.
29. Peng S, Lacerda AE, Kirsch GE, Brown AM, Bruening-Wright A. The action potential
and comparative pharmacology of stem cell-derived human cardiomyocytes.
J Pharmacol Toxicol Methods 2010;61:277 –86.
30. Mordwinkin NM, Burridge PW, Wu JC. A review of human pluripotent stem cellderived cardiomyocytes for high-throughput drug discovery, cardiotoxicity
screening, and publication standards. J Cardiovasc Transl Res 2013;6:22 –30.
31. Raschi E, Ceccarini L, De Ponti F, Recanatini M. hERG-related drug toxicity and
models for predicting hERG liability and QT prolongation. Expert Opin Drug
Metab Toxicol 2009;5:1005 –21.
32. Wallis RM. Integrated risk assessment and predictive value to humans of nonclinical repolarization assays. Br J Pharmacol 2010;159:115 –21.
33. Committee for Medicinal Products for Human Use. ICH topic E14: the clinical
evaluation of QT/QTc interval prolongation and proarrhythmic potential for nonantiarrhythmic drugs questions and answers (EMA/CHMP/ICH/310133/2008).
EMEA, London, May 2012.
34. Bouvy JC, Koopmanschap MA, Shah RR, Schellekens H. The cost-effectiveness of
drug regulation: the example of thorough QT/QTc studies. Clin Pharmacol Ther
2012;91:281 – 8.
35. Fletcher K, Shah RR, Thomas A, Tobin F, Rodriguez B, Mirams GR et al. Novel
approaches to assessing cardiac safety—proceedings of a workshop regulators, industry and academia discuss the future of in silico cardiac modelling to predict the
proarrhythmic safety of drugs. Drug Saf 2011;34:439 – 43.
36. Fink M, Noble D. Pharmacodynamic effects in the cardiovascular system: the modeller’s view. Basic Clin Pharmacol Toxicol 2010;106:243 –9.
37. Mirams GR, Davies MR, Cui Y, Kohl P, Noble D. Application of cardiac electrophysiology simulations to pro-arrhythmic safety testing. Br J Pharmacol 2012;167:
932 –45.
38. Amanfu RK, Saucerman JJ. Cardiac models in drug discovery and development:
a review. Crit Rev Biom Eng 2011;39:379 –95.
39. Brennan T, Fink M, Rodriguez B. Multiscale modelling of drug-induced effects on
cardiac electrophysiological activity. Eur J Pharm Sci 2009;36:62 –77.
40. Rodriguez B, Burrage K, Gavaghan D, Grau V, Kohl P, Noble D. The systems biology
approach to drug development: application to toxicity assessment of cardiac drugs.
Clin Pharmacol Ther 2010;88:130–4.
41. Stockbridge N. Can the Thorough QT Study Be Replaced? CSRC Annual Meeting
2012 & CRF Open Forum. 2012 https://www.cardiac-safety.org/think-tanks/
december-2012.
42. Sahah RR. Drug-induced QT interval prolongation—regulatory guidance and perspectives on hERG channel studies. In Derek J, Goode C, Goode J (eds). The hERG
Cardiac Potassium Channel: Structure, Function and Long QT Syndrome. Chichester:
Novartis Foundation. John Wiley & Sons; 2005.
43. Witchel H, Milnes J, Mitcheson J, Hancox J. Troubleshooting problems with in vitro
screening of drugs for QT interval prolongation using HERG K+ channels
expressed in mammalian cell lines and Xenopus oocytes. J Pharmacol Toxicol
Methods 2002;48:65 –80.
44. Di Veroli GY, Davies MR, Zhang H, Abi-Gerges N, Boyett MR. High-throughput
screening of drug-binding dynamics to HERG improves early drug safety assessment. Am J Physiol Heart Circ Physiol 2013;304:H104 – 17.
45. Kirsch GE, Trepakova ES, Brimecombe JC, Sidach SS, Erickson HD, Kochan MC
et al. Variability in the measurement of hERG potassium channel inhibition:
effects of temperature and stimulus pattern. J Pharmacol Toxicol Methods 2004;50:
93 –101.
734
46. O’Hara T, Rudy Y. Quantitative comparison of cardiac ventricular myocyte electrophysiology and response to drugs in human and nonhuman species. Am J Physiol
Heart Circ Physiol 2012;302:H1023 –30.
47. Sale H, Wang J, O’Hara TJ, Tester DJ, Phartiyal P, He JQ et al. Physiological properties of hERG 1a/1b heteromeric currents and a hERG 1b-specific mutation associated with long-QT syndrome. Circ Res 2008;103:81–95.
48. Vandenberg JI, Varghese A, Lu Y, Bursill JA, Mahaut-Smith MP, Huang CL. Temperature dependence of ether-a-go-go-related gene K+ currents. Am J Physiol Cell
Physiol 2006;291:C165 –75.
49. Zhou Z, Gong Q, Ye B, Fan Z, Makielski JC, Robertson GA et al. Properties of HERG
channels stably expressed in HEK 293 cells studied at physiological temperature.
Biophys J 1998;74:230 –41.
50. Berecki G, Zegers JG, Verkerk AO, Bhuiyan ZA, de Jonge B, Veldkamp MW et al.
HERG channel (dys)function revealed by dynamic action potential clamp technique. Biophys J 2005;88:566 –78.
51. Quinn TA, Granite S, Allessie MA, Antzelevitch C, Bollensdorff C, Bub G et al.
Minimum Information about a Cardiac Electrophysiology Experiment (MICEE):
standardised reporting for model reproducibility, interoperability, and data
sharing. Prog Biophys Mol Biol 2011;107:4–10.
52. Bains W, Basman A, White C. HERG binding specificity and binding site structure:
evidence from a fragment-based evolutionary computing SAR study. Prog Biophys
Mol Biol 2004;86:205 –33.
53. Wiśniowska B, Polak S. hERG in vitro interchange factors—development and verification. Toxicol Mech Methods 2009;19:278 –84.
54. McAllister RE, Noble D, Tsien RW. Reconstruction of the electrical activity of
cardiac Purkinje fibers. J Physiol 1975;251:1 –59.
55. Cherry EM, Fenton FH. A tale of two dogs: analyzing two models of canine ventricular electrophysiology. Am J Physiol Heart Circ Physiol 2007;292:H43 –55.
56. Beeler GW, Reuter H. Reconstruction of the action potential of ventricular myocardial fibres. J Physiol 1977;268:177 – 210.
57. Luo CH, Rudy Y. A model of the ventricular cardiac action potential. Depolarization, repolarization, and their interaction. Circ Res 1991;68:1501 –26.
58. Priebe L, Beuckelmann DJ. Simulation study of cellular electric properties in heart
failure. Circ Res 1998;82:1206 –23.
59. Niederer SA, Fink M, Noble D, Smith NP. A meta-analysis of cardiac electrophysiology computational models. Exp Physiol 2009;94:486 –95.
60. Ten Tusscher KHWJ, Noble D, Noble PJ, Panfilov AV. A model for human ventricular tissue. Am J Physiol 2004;286:1573 –89.
61. Iyer V, Mazhari R, Winslow RL. A computational model of the human leftventricular epicardial myocyte. Biophys J 2004;87:1507 – 25.
62. Heijman J, Volders PG, Westra RL, Rudy Y. Local control of b-adrenergic stimulation: effects on ventricular myocyte electrophysiology and Ca(2+)-transient. J Mol
Cell Cardiol 2011;50:863 –71.
63. Britton OJ, Bueno-Orovio A, Van Ammel K, Lu HR, Towart R, Gallacher DJ et al.
Experimentally calibrated population of models predicts and explains intersubject
variability in cardiac cellular electrophysiology. Proc Natl Acad Sci USA 2013;110:
E2098–105.
64. Sarkar AX, Christini DJ, Sobie EA. Exploiting mathematical models to illuminate
electrophysiological variability between individuals. J Physiol 2012;590(Pt 11):
2555 –67.
65. Walmsley J, Rodriguez JF, Mirams GR, Burrage K, Efimov IR, Rodriguez B. mRNA
expression levels in failing human hearts predict cellular electrophysiological remodeling: a population-based simulation study. PLoS ONE 2013;8:e56359.
66. Polak S, Fijorek K. Inter-individual variability in the pre-clinical drug cardiotoxic
safety assessment—analysis of the age –cardiomyocytes electric capacitance dependence. J Cardiovasc Transl Res 2012;5:321 –32.
67. Polak S, Fijorek K, Glinka A, Wisniowska B, Mendyk A. Virtual population generator
for human cardiomyocytes parameters: in silico drug cardiotoxicity assessment.
Toxicol Mech Methods 2012;22:31 –40.
68. Nattel S, Frelin Y, Gaborit N, Louault C, Demolombe S. Ion-channel
mRNA-expression profiling: insights into cardiac remodeling and arrhythmic substrates. J Mol Cell Cardiol 2010;48:96 –105.
69. Fitzgerald PT, Ackerman MJ. Drug-induced torsades de pointes: the evolving role of
pharmacogenetics. Heart Rhythm 2005;2:S30 –7.
70. Stockbridge N, Morganroth J, Shah RR, Garnett C. Dealing with global safety issues:
was the response to QT-liability of non-cardiac drugs well coordinated? Drug Saf
2013;36:167 – 82.
71. Dickinson GL, Lennard MS, Tucker GT, Rostami-Hodjegan A. The use of mechanistic DM-PK-PD modelling to assess the power of pharmacogenetic studies—
CYP2C9 and warfarin as an example. Br J Clin Pharmacol 2007;64:14–26.
72. Clancy CE, Zhu ZI, Rudy Y. Pharmacogenetics and anti-arrhythmic drug therapy: a
theoretical investigation. Am J Physiol Heart Circ Physiol 2007;292:H66 –75.
73. Ackerman MJ, Tester DJ, Jones GS, Will ML, Burrow CR, Curran ME. Ethnic differences in cardiac potassium channel variants: implications for genetic susceptibility
B. Wiśniowska et al.
74.
75.
76.
77.
78.
79.
80.
81.
82.
83.
84.
85.
86.
87.
88.
89.
90.
91.
92.
93.
94.
95.
96.
97.
98.
99.
to sudden cardiac death and genetic testing for congenital long QT syndrome. Mayo
Clin Proc 2003;78:1479 –87.
Anson BD, Ackerman MJ, Tester DJ, Will ML, Delisle BP, Anderson CL et al. Molecular and functional characterization of common polymorphisms in HERG
(KCNH2) potassium channels. Am J Physiol Heart Circ Physiol 2004;286:H2434 –41.
Arnestad M, Crotti L, Rognum TO, Insolia R, Pedrazzini M, Ferrandi C et al. Prevalence of long-QT syndrome gene variants in sudden infant death syndrome. Circulation 2007;115:361–7.
Bezzina CR, Verkerk AO, Busjahn A, Jeron A, Erdmann J, Koopmann TT et al.
A common polymorphism in KCNH2 (HERG) hastens cardiac repolarization. Cardiovasc Res 2003;59:27 –36.
He FZ, McLeod HL, Zhang W. Current pharmacogenomic studies on hERG potassium channels. Trends Mol Med 2013;19:227 –38.
Fink M, Noble D, Virag L, Varro A, Giles WR. Contributions of HERG K+ current
to repolarization of the human ventricular action potential. Prog Biophys Mol Biol
2008;96:357 –76.
Larsen AP, Olesen S-O. Differential expression of hERG1 channel isoforms reproduces properties of native IKr and modulates cardiac action potential characteristics. PLoS ONE 2010;5:e9021.
Jons C, O-Uchi J, Moss AJ, Reumann M, Rice JJ, Goldenberg I et al. Use of mutantspecific ion channel characteristics for risk stratification of long QT syndrome
patients. Sci Transl Med 2011;3:76ra28.
Hoefen R, Reumann M, Goldenberg I, Moss AJ, O-Uchi J, Gu Y et al. In silico cardiac
risk assessment in patients with long QT syndrome. Type 1: clinical predictability of
cardiac models. J Am Coll Cardiol 2012;60:2182 – 91.
Song W, Xiao Y, Chen H, Ashpole NM, Piekarz AD, Ma P et al. The human Nav1.5
F1486 deletion associated with long QT syndrome leads to impaired
sodium channel inactivation and reduced lidocaine sensitivity. J Physiol 2012;590:
5123 –39.
Martin RL, McDermott JS, Salmen HJ, Palmatier J, Cox BF, Gintant GA. The utility of
hERG and repolarization assays in evaluating delayed cardiac repolarization: influence of multi-channel block. J Cardiovasc Pharmacol 2004;43:369–79.
Hoffmann P, Warner B. Are hERG channel inhibition and QT interval prolongation
all there is in drug induced torsadogenesis? A review of emerging trends. J Pharmacol
Toxicol Methods 2006;53:87 –105.
Lacerda AE, Kuryshev YA, Chen Y, Renganathan M, Eng H, Danthi SJ et al. Alfuzosin
delays cardiac repolarization by a novel mechanism. J Pharmacol Exp Ther 2008;324:
427 –33.
Rosen MR, Ilvento JP, Gelband H, Merker C. Effects of verapamil on electrophysiologic properties of canine cardiac Purkinje fibers. J Pharmacol Exp Ther 1974;189:
414 –22.
Towart R, Linders JT, Hermans AN, Rohrbacher J, van der Linde HJ, Ercken M et al.
Blockade of the I(Ks) potassium channel: an overlooked cardiovascular liability in
drug safety screening? J Pharmacol Toxicol Methods 2009;60:1– 10.
Chadwick D, Goode J. Novartis Foundation. The hERG Cardiac Potassium Channel:
Structure, Function, and Long QT Syndrome. New York: John Wiley; 2005.
Polak S, Wiśniowska B, Fijorek K, Glinka A, Polak M, Mendyk A. Modelling and simulation approach for assessing proarrhythmic potency of the non-cardiological
drugs. Comput Cardiol 2012;39:761 – 4.
Kramer J, Obejero-Paz CA, Myatt G, Kuryshev YA, Bruening-Wright A, Verducci JS
et al. MICE models: superior to the HERG model in predicting Torsade de Pointes.
Sci Rep 2013;3:2100.
Polak S. In vitro to human in vivo translation—pharmacokinetics and pharmacodynamics of quinidine. ALTEX 2013;3:309 –18.
Starmer CF. Theoretical characterization of ion channel blockade. Competitive
binding to periodically accessible receptors. Biophys J 1987;52:405 –12.
Starmer CF, Courtney KR. Modelling ion channel blockade at guarded binding sites:
application to tertiary drugs. Am J Physiol 1986;251(Pt 2):H848 – 56.
Starmer CF, Grant AO. Phasic ion channel blockade. A kinetic model and parameter estimation procedure. Mol Pharmacol 1985;28:348 –56.
Starmer CF, Grant AO, Strauss HC. Mechanisms of use-dependent block of
sodium channels in excitable membranes by local anesthetics. Biophys J 1984;46:
15 –27.
Michalets EL, Williams CR. Drug interactions with cisapride: clinical implications.
Clin Pharmacokinet 2000;39:49 –75.
Friemel A, Zünkler BJ. Interactions at human ether-à-go-go-related gene channels.
Toxicol Sci 2010;114:346 –55.
Lu HR, Hermans AN, Gallacher DJ. Does terfenadine-induced ventricular tachycardia/fibrillation directly relate to its QT prolongation and Torsades de Pointes?
Br J Pharmacol 2012;166:1490 – 502.
Miglioli PA, Merlo F, Fabbri A, Padrini R. Teicoplanin concentrations in serum, pericardium, pericardial fluid and thoracic wall fat in patients undergoing cardiopulmonary bypass surgery. J Antimicrob Chemother 1997;39:229 –33.
Challenges with computer-based prediction of the drug proarrhythmic effect
100. Beam TR Jr, Galask RP, Friedhoff LT, Platt TB, Leitz MA. Aztreonam concentrations
in human tissues obtained during thoracic and gynecologic surgery. Antimicrob
Agents Chemother 1986;30:505.
101. van der Linde HJ, Van Deuren B, Somers Y, Loenders B, Towart R, Gallacher DJ. The
electro-mechanical window: a risk marker for Torsade de Pointes in a canine model
of drug induced arrhythmias. Br J Pharmacol 2010;161:1444 –54.
102. Guns PJ, Johnson DM, Weltens E, Lissens J. Negative electro-mechanical windows
are required for drug-induced Torsades de Pointes in the anesthetized guinea pig.
J Pharmacol Toxicol Methods 2012;66:125 – 34.
103. Williams GS, Smith GD, Sobie EA, Jafri MS. Models of cardiac excitation–contraction
coupling in ventricular myocytes. Math Biosci 2010;226:1–15.
104. Gintant G. Ions, equations and electrons: the evolving role of computer simulations
in cardiac electrophysiology safety evaluations. Br J Pharmacol 2012;167:929 –31.
105. Mirams GR, Noble D. Is it time for in silico simulation of drug cardiac side effects? Ann
N Y Acad Sci 2011;1245:44 –7.
735
106. Moreno JD, Clancy CE. Using computational modeling to predict arrhythmogenesis and antiarrhythmic therapy. Drug Discov Today Dis Models 2009;6:71–84.
107. Noble D. Computational models of the heart and their use in assessing the actions
of drugs. J Pharmacol Sci 2008;107:107 – 17.
108. Soubret A, Helmlinger G, Dumotier B, Bibas R, Georgieva A. Modeling and simulation of preclinical cardiac safety: towards an integrative framework. Drug Metab
Pharmacokinet 2009;24:76–90.
109. Valerio LG Jr, Balakrishnan S, Fiszman ML, Kozeli D, Li M, Moghaddam S et al.
Development of cardiac safety translational tools for QT prolongation and
torsade de pointes. Expert Opin Drug Metab Toxicol 2013;9:801 – 15.
110. Redfern WS, Carlsson L, Davis AS, Lynch WG, MacKenzie I, Palethorpe S et al.
Relationships between preclinical cardiac electrophysiology, clinical QT
interval prolongation and torsade de pointes for a broad range of drugs: evidence for a provisional safety margin in drug development. Cardiovasc Res
2003;58:32 – 45.