Electrophysiological and ECG Effects of
Perhexiline, a Mixed Cardiac Ion Channel
Inhibitor, Evaluated in Nonclinical Assays and
in Healthy Subjects
Mark G. Midei, MD1, Borje Darpo, MD, PhD2, Greg Ayers, MD, PhD1,
Randy Brown, MSc2, Jean-Philippe Couderc, PhD2,William Daly, JD1,
Georg Ferber, PhD3, Philip T. Sager, MD4, and A. John Camm, MD5
Abstract
Perhexiline has been used to treat hypertrophic cardiomyopathy. In addition to its effect on carnitine-palmitoyltransferase-1, it has mixed ion channel
effects through inhibition of several cardiac ion currents. Effects on cardiac ion channels expressed in mammalian cells were assayed using a manual
patch-clamp technique, action potential duration (APD) was measured in ventricular trabeculae of human donor hearts, and electrocardiogram effects
were evaluated in healthy subjects in a thorough QT (TQT) study. Perhexiline blocked several cardiac ion currents at concentrations within the
therapeutic range (150-600 ng/mL) with IC50 for hCav1.2 ∼ hERG < late hNav1.5. A significant APD shortening was observed in perhexiline-treated
cardiomyocytes. The TQT study was conducted with a pilot part in 9 subjects to evaluate a dosing schedule that would achieve therapeutic and
supratherapeutic perhexiline plasma concentrations on days 4 and 6, respectively. Guided by the results from the pilot, 104 subjects were enrolled
in a parallel-designed part with a nested crossover comparison for the positive control. Perhexiline caused QTc prolongation, with the largest effect
on QTcF, 14.7 milliseconds at therapeutic concentrations and 25.6 milliseconds at supratherapeutic concentrations and a positive and statistically
significant slope of the concentration-QTcF relationship (0.018 milliseconds per ng/mL; 90%CI, 0.0119-0.0237 milliseconds per ng/mL). In contrast,
the JTpeak interval was shortened with a negative concentration-JTpeak relationship, a pattern consistent with multichannel block. Further studies
are needed to evaluate whether this results in a low proarrhythmic risk.
Keywords
CiPA, concentration-effect modeling, JTpeak, multi-channel block, perhexiline, QT
Perhexiline is a carnitine-palmitoyltransferase 1 (CPT-
1) inhibitor that has been used to treat symptomatic
hypertrophic cardiomyopathy (HCM).1–3 The drug was
developed in the 1970s as an anti-ischemic drug for
treatment of angina pectoris and has been used clini￾cally for this indication in Australia and New Zealand
for >30 years. Compared with other agents available
at the time of its introduction, perhexiline had the
unusual property of not inducing intolerable reductions
in blood pressure or heart rate.4 Inhibition of CPT-1
causes a change in the preferred substrate for energy
production in cardiac myocytes from fatty acids to
glucose, an inherently more oxygen-efficient process.
This mechanism, in combination with its cardiac ion
channel-blocking properties, may reverse the impaired
cardiac energetics present in HCM associated with ab￾normal myocyte function.2 Recognizing the role of cel￾lular ischemia in the pathophysiology of HCM and the
high risk of pharmacologically induced hemodynamic
fluctuations, perhexiline has been used increasingly for
this condition.4
Perhexiline is known to have inhibitory effects
on multiple cardiac ion channels,5,6 including the
human ether-à-go-go-related gene (hERG) channel.7
Although it is known to prolong the QTc interval
on an electrocardiogram (ECG), only a single
case of torsades de pointes has been reported in
more than 40 years of clinical experience.8 Others
have reported favorable suppression of cardiac
arrhythmias.9–11 To fully characterize the cardiac
1Heart Metabolics, Dublin, Ireland
2ERT, Rochester, New York, USA
3Statistik Georg Ferber GmbH, Riehen, Switzerland
4Department of Medicine, Cardiovascular Research Institute, Stanford
University, Palo Alto, California, USA
5Division of Clinical Sciences,Cardiovascular and Cell Sciences Research
Institute, St George’s University of London, London, UK
Submitted for publication 25 April 2021; accepted 28 June 2021.
Corresponding Author:
Borje Darpo, MD, PhD, Chief Scientific Officer Cardiac Safety, ERT, 150
Allens Creek Road, Rochester, NY 14618
Email: [email protected]
2 The Journal of Clinical Pharmacology / Vol 0 No 0 2021
ion channel effects of perhexiline and its association
with its electrophysiological properties, the effects of
perhexiline on cardiac ion currents, action potential
duration (APD) of human heart, and ECG parameters
were measured including concentration (C)-QTc and
C-JTpeak evaluation in healthy human subjects.
Methods
Effects of Perhexiline on Cardiac Ion Channels
The effect of perhexiline on the following ion channels
expressed in mammalian cells using manual patch￾clamp techniques was examined (Charles River, Inc.,
Cleveland, Ohio):
1. The sodium channel (hNav1.5) responsible for
peak and late INa in human heart, expressed in
human embryonic kidney (HEK293) cells.
2. The l-type calcium channel (hCav1.2-β2-α2δ) re￾sponsible for ICa,l in human heart, expressed in
Chinese hamster ovary cells.
3. The potassium channel (hKvLQT1/hminK) re￾sponsible for the slow delayed rectifier current
(IKs) in human heart, expressed in HEK293 cells.
4. The hERG potassium channel current (a sur￾rogate for IKr, the rapidly activating, delayed
rectifier cardiac potassium current) expressed in
HEK293 cells. This assay was conducted in com￾pliance with Good Laboratory Practice (GLP).
A manual patch-clamp technique at near￾physiological temperature was used for all channels
except Nav1.5, which was evaluated at room
temperature. At least 44 concentrations were selected to
evaluate the concentration-response of each channel,
except for hERG, which was evaluated at1 a single
concentration. Each concentration was tested in at
least 33 cells (n ≥ 3), with one exception, in which a
10-μM concentration of perhexiline was applied to
only 2 cells expressing Cav1.2 current. Because this
was determined to be a saturating concentration, 4
lower concentrations were subsequently tested. Steady
state was maintained for at least 30 seconds before
applying the test article or the positive control. The
peak current during the test article and positive control
application was monitored until a new steady state was
achieved. A different procedure was followed for the
hERG assay. The test article was applied for a period
of ≈5 minutes and then washed off. Monitoring of
hERG current was then continued for up to 15 minutes
to determine whether perhexiline’s effect on hERG
was attenuated. Samples of the test article formulation
solutions collected from the outflow of the perfusion
apparatus were analyzed for concentration verification
during the hERG assay.
Effects of Perhexiline on Human Action Potential Dura￾tion
Trabeculae were harvested from 7 whole human donor
hearts (5 from healthy subjects and 2 from subjects with
left ventricular hypertrophy) procured by AnaBios (San
Diego, California). Hearts were treated with a cardio￾plegic solution for transit. After ventricular trabeculae
were prepared, continuous membrane potential mon￾itoring was performed in the presence of oxygenated
Tyrode’s buffer.
Specimens were studied before and after exposure
to 3 different concentrations of perhexiline: 0.72, 2.2,
and 7.2 μM (200, 610, and 2000 ng/mL). The fol￾lowing experimental parameters were varied to assess
the effect of perhexiline under a variety of conditions:
(1) different vehicles/carriers were used—3.5% human
serum albumin, 0.3% human serum albumin, or 0.1%
dimethylsulfoxide; (2) different drug exposure times
were used, from 30 minutes to 4 hours; (3) 2 differ￾ent pacing frequencies were tested, 1 and 2 Hz; and
(4) normal as well as pathological (left ventricular
hypertrophy) hearts were used.
Dofetilide at a concentration of 100 nM was used
as a positive control after data were collected for the
perhexiline exposures. Analysis included APD at 30,
60, and 90 minutes (APD30,60,90), resting membrane
potential (RMP), maximum amplitude of action poten￾tial (APA), short-term variability analysis (STV), and
triangulation (APD90 −APD30).
ECG Effects in Healthy Subjects—The Thorough QT
Study
Regulatory approval was granted by the Therapeutic
Goods Administration in Australia. Ethics committee
approval was granted by the Belberry Human Research
Ethics Committee. The study was conducted in accor￾dance with the current version of the Declaration of
Helsinki (52nd WMA General Assembly, Edinburgh,
Scotland; October 2000). The trial was conducted in
agreement with the International Conference on Har￾monization (ICH) guidelines on Good Clinical Prac￾tice. All subjects provided written informed consent to
participate in the study prior to being screened.
The thorough QT (TQT) study was conducted
in compliance with ICH E14 guidance12,13 and was
double-blind, randomized, and placebo-controlled
with a parallel-group design and used a nested
crossover comparison for the test of assay sensitivity
with moxifloxacin.14 The study was preceded by a
small pilot phase, in which 9 subjects were dosed with
perhexiline to evaluate whether the proposed dosing
scheme resulted in the targeted plasma concentrations,
that is, 150 to 400 ng/mL after 4 days of dosing
with the proposed schedule and somewhat above
600 ng/mL after 6 days. Subjects were randomly
Midei et al 3
assigned 1:1 to 1 active group (group 1) and 1
placebo/moxifloxacin group (group 2). The target
for the perhexiline dosing schedule was to achieve
therapeutic plasma concentrations on day 4 and
supratherapeutic concenrtrations on day 6. Perhexiline
was given in a fixed dosing of 200 mg orally twice
daily on days 1 and 2, once daily on days 3 and 4,
and a single dose between 150 and 300 mg on day 5,
guided by a measurement of plasma concentration,
and then a single dose of 400 mg on day 6. In the active
group, subjects received placebo for moxifloxacin
on day −1 and day 7. In the control group, subjects
received placebo for perhexiline on days 1 through 6.
In half the group (n = 25), subjects received a single
dose of 400 mg moxifloxacin on day −1 and placebo
for moxifloxacin on day 7; in the other half (n =
25), placebo and moxifloxacin were administered in
reversed order with moxifloxacin on day 7.
Subjects were admitted to the clinical site (Nucleus
Network, Melbourne, Australia) 3 days before dosing
(day −3) and stayed in-house until day 8, after comple￾tion of safety procedures. Standard eligibility criteria
for clinical pharmacology studies in healthy subjects
were applied. Cytochrome P-450 (CYP) 2D6 poor me￾tabolizers and ultrarapid metabolizers as determined
via genotype were excluded from participation.
ECG Technique and Statistical Analyses. Twelve-lead
ECGs were extracted from continuous Holter record￾ings (M12R 12-lead digital ECG recorders; Global
Instrumentation, Manlius, New York) on days −2, −1,
4, 6, and 7 at predefined times intended to capture
peak plasma levels of perhexiline and moxifloxacin
and declining levels thereafter, with subjects supinely
resting for at least 10 minutes. ECGs were extracted
and measured at a central ECG laboratory (iCardiac
Technologies/ERT, Inc., Rochester, New York). A me￾dian beat was computed from the vector magnitude
lead,8,13 derived from 12-lead ECGs with Mason-Likar
lead configuration,14 and transformed into a 3-lead
orthogonal configuration (X, Y, and Z) by implement￾ing the Guldenring procedure.15 The magnitude of the
vector described by these 3 leads was computed as the
root of the sum of squares, and finally a median beat
was computed by synchronizing all sinus beats within
the 10-second replicate (by maximization of cross￾correlation function on QRS complexes).16 Fiducial
points were detected on each replicate median beat us￾ing iCOMPAS software and were manually adjudicated
by the technician/cardiologist. Beats were defined as
usable when they were both (1) normal sinus beats and
(2) the correct placement of the fiducials of interest
could be accurately determined. The ECG technicians
were trained to use consistent rules in defining the peak
of the T wave. In case of a biphasic T wave, the T-wave
peak was based on the peak of the major inflection of
the T wave (in absolute value). On each median beat, the
technician reviewed the automated placement of the P
onset, QRS onset and offset, T apex, and T offset. If
the automated fiducial placement was determined to be
within 10 milliseconds of the correct placement it was
left unchanged; otherwise, the fiducials were adjusted as
needed. Subsequently, the JTpeak measurements were
reviewed and adjudicated by a trained cardiologist. The
QT interval was corrected for heart rate changes using
the Fridericia formula (QTcF) and the JTpeak interval
using a larger correction factor of 0.58 (JTpeak_c =
JT × RR−0.58), as suggested by Johannesen.17
The primary end point in the statistical analy￾sis was placebo-corrected change-from-baseline QTcF
(QTcF) with heart rate, PR, QRS, T-wave mor￾phology and JTpeak and Tpeak-Tend as secondary end
points. For the evaluation of the ECG effect of per￾hexiline, baseline was time-matched values on day −2.
For the nested crossover comparison between placebo
and moxifloxacin, baseline was the time-matched values
on either day −1 or day 7. The primary analysis was
performed separately for day 4 and day 6. It was
based on a linear mixed-effects model with change￾from-baseline QTcF (QTcF) as dependent variable,
treatment (perhexiline and corresponding placebo),
time (categorical), and treatment by time as factors.
Subject was included in the model as a random effect for
the intercept. An unstructured covariance matrix was
specified for the repeated measures at postdose times
within subject. A 2-sided 90%CI was calculated for the
contrast in treatment QTc = “perhexiline-placebo.” For
JTpeak and Tpeak-Tend analysis, the same model as for
QTcF was used.
The analysis to show assay sensitivity was based
on the change in QTc (primary end point) from
time-matched baseline evaluated on moxifloxacin
and matching placebo for group 2. For each time, a
linear mixed-effects model was fitted with treatment
(moxifloxacin and corresponding placebo) and
sequence (placebo-moxifloxacin or moxifloxacin￾placebo) as fixed effects, time-matched baseline as
a covariate, and subject as a random effect for the
intercept. For times 1, 2, 3, and 4 hours, the contrast in
QTc = “moxifloxacin-placebo” was tested against the
1-sided null hypothesis QTc ≤ 5 milliseconds at the
5% level. Multiplicity was controlled using a Hochberg
procedure.18 If, following this procedure, the QTc was
significantly > 5 milliseconds for at least 1 point, assay
sensitivity was considered to be shown.
The analysis of placebo-corrected HR, PR, and
QRS was based on the same linear mixed-effects
model as for QTcF and JTpeak_c.
In the concentration-QTc analysis, a plot of stan￾dardized residuals versus fitted values was used to
4 The Journal of Clinical Pharmacology / Vol 0 No 0 2021
Table 1. Perhexiline IC50 for Cardiac Ion Channels Using the Manual
Patch-Clamp Technique
Ion Channel (±)-Perhexiline
IC50 (ng/mL)
Manual Patch
Cell Type
hERG (non-GLP) 1139 HEK293
hERG (GLP)a 167 HEK293
Late hNav1.5 1359 HEK293
hNav1.5 tonic 1324 HEK293
hNav1.5 phasic 1069 HEK293
hCav1.2 247 CHO
hKvLQT1/hminK 1212 HEK293
hKv1.5 2455 CHO
hKv4.3/hKChip2.2 hKv4.3/hKChip 5874 HEK293
hKir2.1 >8325 HEK293
HEK293, Human embryonic kidney cells; CHO, Chinese hamster ovary cells.
GLP compliant.
examine departure from assumptions of a linear model.
The normal Q-Q plots of the random effects and
the within-subject errors were used to investigate the
normality of the random effects and the within-subject
errors, respectively. A final assessment of the adequacy
of the linear mixed-effects model was provided by a
goodness-of -fit plot (ie, the observed concentration
decile-QTcF plot). Via visual inspection of the
goodness-of -fit plot, the assumption of linearity be￾tween QTcF and plasma concentrations of perhex￾iline and how well the predicted QTcF matched the
observed data in the regions of interest were checked.
Plots over QTcF and perhexiline concentrations
over time and hysteresis loops were used to evaluate the
presence of hysteresis, that is, a relevant delay between
peak plasma concentration and the largest QT effect. A
linear mixed-effects model was applied to the data for
the C-QTc analysis of the relationship between perhex￾iline plasma concentration and QTcF/JTpeak_c with
separate analyses for each ECG parameter. The time￾matched concentration of perhexiline was included in
the model as a continuous covariate (0 for placebo),
with centered baseline QTcF/JTpeak_c as an additional
covariate and treatment (active or placebo) and time as
categorical factors; there was a random intercept and
slope per subject. Degrees of freedom for the model es￾timates was determined by the Kenward Roger method.
From the model, the slope and the treatment effect￾specific intercept (defined as the difference between
active and placebo) were estimated together with a
2-sided 90%CI. The predicted QTcF/JTpeak_c
effect and 2-sided 90%CI at the geometric mean Cmax
of perhexiline were calculated.
Results
Effects of Perhexiline on Cardiac Ion Channels
Table 1 shows the perhexiline concentrations required
to inhibit 50% of the respective cardiac ion channels
Figure 1. Representative action potential duration for LVH specimen
stimulated at 1 Hz. Shortening of the action potential duration (APD)
was seen consistently at all concentrations of perhexiline compared with
dofetilide administered as a control.
(IC50). Perhexiline inhibition of the l-type calcium
current was near equipotent to the hERG inhibition
(IC50, 247 vs 167 ng/mL), whereas 50% inhibition
of the late sodium current was achieved at 8-fold
higher concentrations than hERG inhibition (IC50,
1359 ng/mL). In the GLP-compliant hERG assay, the
perhexiline effect plateaued at 66% inhibition, with
subsequent rapid but incomplete recovery on washout
to 48% inhibition, which remained for another 9 to
15 minutes. At the upper therapeutic concentration of
600 ng/mL, perhexiline is ∼99% protein bound, which
means that the ratio between IC50 and free therapeutic
concentration was ∼28-fold for hERG and ∼40-fold for
the l-type calcium current.
Effects of Perhexiline on the Human Action Potential
Perhexiline at the studied concentrations up to 2000
ng/mL had an overall shortening effect on APD and
no effect on triangulation in contrast to dofetilide,
administered as a control after data were obtained for
perhexiline. Action potential results from a represen￾tative left ventricular hypertrophy (LVH) specimen is
displayed for the control condition, for perhexiline at
various concentrations, and for dofetilide in Figure 1.
Average percentage change obtained in 5 specimens
derived from 2 of the LVH hearts for APD, APA, and
RMP maximum dV/dt and for triangulation is shown
in Table 2. Under the conditions tested, a consistent
shortening of the action potentials was observed. There
were no increases in action potential triangulation,
action potential instability or early afterdepolarization.
The shortening of the APD was associated with a small,
Midei et al 5
Table 2. Effects of Perhexiline on Action Potential Parameters (Mean Change, %) for the 5 Samples Obtained From the 2 Human Donors With LVH
APD30 (%) APD50 (%) APD60 (%) APD90 (%)
Control 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00
Perhexiline 200 ng/mL −3.33 ± 0.78 −3.36 ± 0.79 −3.14 ± 0.81 −0.72 ± 1.97
Perhexiline 610 ng/mL −6.93 ± 2.76 −5.92 ± 2.44 −5.52 ± 2.31 −7.13 ± 3.50
Perhexiline 2000 ng/mL −10.23 ± 2.65 −8.75 ± 2.31 −7.94 ± 2.17 −7.62 ± 2.57
Dofetilide 44 ng/mL 1.35 ± 3.42 11.57 ± 4.23 16.99 ± 4.75 32.87 ± 6.75
APA (%) RMP (%) Max dV/dt (%) Triangulation APD90-30 (%)
Control 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00 0.00 ± 0.00
Perhexiline 200 ng/mL 0.00 ± 1.51 0.77 ± 1.32 55.08 ± 47.65 −1.34 ± 1.30
Perhexiline 610 ng/mL −0.40 ± 3.37 3.83 ± 1.74 22.84 ± 45.8 0.32 ± 2.29
Perhexiline 2000 ng/mL −4.18 ± 3.58 0.92 ± 2.62 −9.68 ± 20.19 2.01 ± 0.89
Dofetilide 44 ng/mL −0.67 ± 3.26 −0.10 ± 4.35 74.72 ± 66.3 85.11 ± 13.52
APD, action potential duration at 30%, 50%, 60%, and 90% repolarization;APA, action potential amplitude;RMP, resting membrane potential;Max dV/dt,membrane
potential velocity; triangulation, APD90 – APD30.
not statistically significant increase of mean STV: 0.22
for control; 0.71, 0.45, and 0.43 for perhexiline 200,
610, and 2000 ng/mL, respectively; and 1.21 for 44
ng/mL dofetilide. In the experiments in which 44 ng/mL
(100 nM) dofetilide was added as a positive control
after the perhexiline challenge, the effect of dofetilide
on APD90 was ≈60% smaller than typically observed,
only 32.9% ± 6.75% over baseline versus 102% (data
on-file, Anabios).
Effects of Perhexiline in the TQT Study
One hundred four subjects (63 men) were enrolled
into the study and received at least 1 dose of study
medication (52 in the perhexiline group and 26 each
in the combined moxifloxacin/placebo groups). There
were no apparent differences between the groups for
demographic parameters of age, sex, race, body weight,
and body mass index. The majority of subjects (87%)
were white with a mean ± SD age and body mass
index of 25.5 ± 5.0 years and 23.3 ± 2.66 kg/m2,
respectively.
Pharmacokinetics of Perhexiline. The dosing schedule
of perhexiline was successful in achieving a mean
plasma level of perhexiline within the therapeutic range
(150-400 ng/mL) on day 4 and somewhat above the
therapeutic range on day 6. Mean Cmax was 279 ng/mL
(90%CI, 240-323 ng/mL) on day 4 and 678 ng/mL
(90%CI, 621-740 ng/mL) on day 6, with a median Tmax
of 6.0 hours on both days (Figure 2).
Cardiodynamic Effects. No significant changes in
blood pressure or heart rate were seen with perhexiline
on day 4 or day 6. The change-from-baseline heart rate
(HR) followed the same diurnal pattern in the per￾hexiline and placebo groups on day 4, whereas a small
separation was observed on day 6, with mean HR
reaching 6.4 and 8.5 bpm in subjects on perhexiline as
compared with 1.5 and 1.5 bpm among placebo subjects
3 and 4 hours postdosing, respectively (Figure S1).
The resulting mean placebo-corrected HR (HR)
peaked at 6.9 bpm 4 hours postdosing on day 6
(Figure S2).
Change-from-baseline QTcF (QTcF) was larger
at all times in the perhexiline group compared with
placebo on both day 4 and day 6. On day 4, mean
QTcF varied across postdosing times between
9.8 and 3 hours postdose and 15.3 milliseconds at
24 hours, whereas the placebo response was slightly
negative. On day 6, the QT effect of perhexiline was
larger than on day 4, with a largest mean QTcF of
26.7 milliseconds 24 hours postdosing (Figure 3a).
Mean placebo-corrected QTcF (QTcF) was
between 10 and 15 milliseconds for all the times on
day 4, with the largest value observed at the predose
time (14.7 milliseconds). On day 6, QTcF was
generally larger, starting with a mean values near
20 milliseconds and then increasing to a peak effect of
25.6 milliseconds (90%CI, 20.47-30.70 milliseconds)
24 hours postdosing (Table S1). Assay sensitivity was
confirmed by the QT response after a single dose of
400 mg moxifloxacin, with the largest mean QTcF
of 11.6 milliseconds (90%CI, 8.53-14.65 milliseconds)
at 4 hours and the lower bound of the 90% CI exceeding
5 milliseconds at all 4 prespecified times: 7.00, 8.50,
6.73, and 8.53 milliseconds 1, 2, 3 and 4 hours
postdosing.
Unlike the QTc interval, JTpeak_c after dosing
with perhexiline was negative or around zero at most
times, except for the 24-hour point on day 6, with
mean JTpeak_c of 7.8 milliseconds (90%CI, 4.47-
11.06 milliseconds; Figure 3b). Mean placebo-corrected
JTpeak_c (JTpeak_c) was below 10 milliseconds
on both day 4 and day 6, with the largest value at
6 The Journal of Clinical Pharmacology / Vol 0 No 0 2021
Figure 2. Plasma concentration profile of perhexiline and metabolites on days 4 and 6. Mean (ng/mL) and standard deviation. PEX, perhexiline; CIS,
cishydroxyperhexiline; TRANS, transhydroxyperhexiline.
2 hours (4.8 milliseconds; 90%CI, 0.93-8.72 millisec￾onds) on day 4 and at the predose time (8.4 millisec￾onds; 90%CI, −0.01 to 16.76 milliseconds) and 24
hours postdose on day 6 (6.5 milliseconds; 90%CI,
1.92-11.17 milliseconds; Table S1). To directly compare
the effect of perhexiline on placebo-corrected QTcF
and JTpeak_c, the profiles across postdosing times
on days 4 and 6 are displayed in the same graph in
Figure 4. In contrast, moxifloxacin, with a smaller mean
peak effect than perhexiline (11.6 milliseconds), caused
prolongation of the JTpeak_c, which peaked at the
same point (4 hours postdosing), amounting to 7.8
milliseconds (90%CI, 4.75-10.80 milliseconds).
There were 4 subjects (7.7%), 3 on day 4 and 1 on
day 6, in the perhexiline group with QTcF > 480 and
< 500 milliseconds at any postdose time versus none in
the placebo and no subject on perhexiline with QTcF >
500 milliseconds versus 1 (1.9%) in the placebo group.
Five subjects (9.6%) in the perhexiline group exhibited
QTcF > 60 milliseconds at total of 11 times versus 1
subject (1.9%) at 1 time in the placebo group.
In the concentration-QTc analysis, plots over
QTcF and perhexiline concentrations over time
(Figure S3) and hysteresis loops did not indicate
the presence of hysteresis, that is, a relevant delay
between peak plasma concentration and the largest
QT effect. The test for linearity demonstrated that a
linear model would be appropriate, and the goodness￾of -fit plot (Figure 5a) showed that the model captured
the observed data across all concentration deciles
in an acceptable way. The slope of the C-QTc
relationship was positive and statistically significant
(0.018 milliseconds per ng/mL; 90%CI, 0.0119-
0.0237 milliseconds per ng/mL), with a large and
statistically significant intercept (10.2 milliseconds;
90%CI, 7.41-13.03 milliseconds). The slope of the C￾JTpeak relationship was negative and not statistically
significant: −0.004 milliseconds per ng/mL (90%CI,
−0.0105 to 0.0025 milliseconds per ng/mL), with an
intercept of 2.7 milliseconds (90%CI, −1.61 to 7.04
milliseconds per ng/mL). The goodness-of -fit plot
indicated that the model underestimated the effect
of perhexiline on the JTpeak_c in the highest
concentration decile (Figure 5b). Using the linear
C-QTc/JTpeak models, the effect on QTcF at
the geometric mean Cmax on day 4 (279 ng/mL)
and on day 6 (678 ng/mL) can be predicted as 15.2
milliseconds (90%CI, 12.67-17.67 milliseconds) and
22.3 milliseconds (90%CI, 18.61-25.93 milliseconds),
respectively. The effect on JTpeak_c at the same
perhexiline concentrations can be predicted as 1.6
milliseconds (90%CI, −2.46 to 5.66 milliseconds) and
0.01 milliseconds (90%CI, −4.93 to 4.95 milliseconds).
Perhexiline exerted a small effect on the PR inter￾val. The largest mean placebo-corrected PR (PR)
reached 7.7 milliseconds (90%CI, 5.08-10.27 millisec￾onds) at 3 hours on day 4 and 9.8 milliseconds (90%CI,
7.09-12.42 milliseconds) at 6 hours on day 6. There were
Midei et al 7
Figure 3. Mean change from time-matched baseline (A) QTcF (QTcF) and (B) JTpeak_c (JTpeak_c) across times on days 4 and 6. Mean and
90%CI from the linear mixed-effects model.
no subjects with a PR > 25% to a value >200 mil￾liseconds. Mean change-from-baseline QRS (QRS)
was somewhat larger on perhexiline on day 6 compared
with placebo, and the resulting mean placebo-corrected
QRS (QRS) reached levels slightly above 2 mil￾liseconds 4, 6, and 8 hours postdose on this day, with
mean QRS of 2.0, 2.1, and 2.4 milliseconds (90%CI,
1.41-3.41 milliseconds), respectively. There were no sub￾jects with QRS > 25% to a value >120 milliseconds
on any of the days.
Safety
The study drugs were safe and well tolerated; there were
no arrhythmias or serious adverse effects reported. Self -
limited symptoms of headache, nausea, and dizziness
occurred in 58% of perhexiline-treated subjects and did
not require discontinuation of study drug.
Discussion
Perhexiline has been used for decades to treat coronary
ischemia and HCM in Australia without significant
8 The Journal of Clinical Pharmacology / Vol 0 No 0 2021
Figure 4. Placebo-corrected change-from-baseline QTcF and JTpeak_c (QTcF and JTpeak_c, milliseconds) across times on day 4 and day 6.
Mean and 90%CI from the linear mixed-effects model.
reports of proarrhythmias. The Australian package
insert notes QT prolongation as a risk, but the level of
this effect has not been well characterized. In this series
of nonclinical and clinical studies, the effects of per￾hexiline on a variety of cardiac ion channels, on action
potential parameters in human cardiomyocytes and on
ECG biomarkers in healthy subjects were investigated.
The objective of a TQT study in healthy subjects
is to obtain a robust evaluation of a drug’s potential
effects on ECG parameters, with a focus on the QTc
interval; therefore, the study is powered to exclude
a small (10–millisecond) effect on the QTc interval.
Based on the results from the study, the level of ECG
monitoring in subsequent patient trials can be decided;
if there are concerning effects on any of the studied
ECG parameters, the results will also be used to exclude
susceptible patients or to provide cautionary statements
in the package insert once the drug is approved. It is
widely agreed that a drug that causes a >20-millisecond
prolongation of the QTc interval may lead to proar￾rhythmias on the basis of delayed repolarization in
susceptible patients, whereas there is an ongoing debate
about to what extent this also applies to drugs with
only a mild effect, for example, between 10 and 20
milliseconds.13 The QT interval seems to be an overly
sensitive biomarker, and there are examples of mildly
QT-prolonging drugs that do not appear to increase
the risk of proarrhythmias. For drugs with only a mild
effect on the QTc interval, it has also been proposed
that lack of JTpeak prolongation, based on mixed ion
channel effects, may mitigate the proarrhythmic risk
of QT prolongation and thereby reduce or eliminate
the need for ECG monitoring in patient trials.19–21
With this series of nonclinical and clinical studies, an
effort was made to better elucidate the relationships
between ion channel effects and the effects on ECG
biomarkers in humans, with a focus on the QT and
JTpeak intervals and to place the findings in the context
of proarrhythmic risk.
Assessment of perhexiline’s effect on cardiac ion
channels confirmed an inhibitory effect on the hERG
channel at concentrations that are clinically relevant.
Additional ion channel effects were also revealed with
inhibition of the l-type calcium channel with an IC50
value similar to hERG and sodium currents, including
the late inward current, at somewhat higher concen￾trations. Perhexiline can therefore be categorized as a
multichannel-blocking drug with an equipotent effect
on l-type calcium and hERG, a profile that has been
suggested to result in low proarrhythmic risk.22,23 A
similar hERG/l-type calcium IC50 relationship exists for
verapamil that can cause QTc prolongation and has
been claimed to have a low risk of torsades de pointes
(TdP).21
In the in vitro APD assay in human cardiomy￾ocytes, no evidence of significant action potential
prolongation, increased triangulation, action potential
instability, or early after depolarization (EAD) were
seen. In several of the experiments, a small shorten￾ing of the action potential was observed. Although
drugs that shorten the ventricular action potential
can lead to beat-to-beat instability and proarrhythmias
Midei et al 9
Figure 5. Goodness-of-fit graphs for the C-QTcF (A) and C-JTpeak_c (B) analysis. The black line with the gray-shaded area shows the mean
predicted effect with 90%CI across the observed range of perhexiline plasma concentrations. The red filled circles with vertical bars denote the
observed placebo-corrected mean QTcF with 90%CI displayed within each plasma concentration decile. The horizontal red line with notches shows
the range of perhexiline concentrations divided into deciles.
(eg, flecainide),24 the shortening induced by perhexiline
was not associated with a significant increase in short￾term variability. Interestingly, in experiments in which
dofetilide was added as a positive control after the per￾hexiline challenge, the prolongation of APD90 typically
seen with dofetilide was reduced by 60%.
In an in vivo toxicology study in equal numbers of
male and female dogs, animals were dosed for 39 weeks
in escalating-dose cohorts of perhexiline (placebo and
5, 15, and 45 mg/kg/d, with 8 animals in each cohort).
ECG measurements were recorded pretreatment (days
−8 to −5) and 1 to 4 hours postdose on day 1, weeks
13, 26, and 39. Changes in ECG parameters were
very small across all dose groups, without indications
of dose-dependent QTc prolongation or shortening
(supplementary material).
10 The Journal of Clinical Pharmacology / Vol 0 No 0 2021
In contrast to the findings of mild APD short￾ening in human cardiomyocytes, the TQT study
clearly demonstrated dose- and concentration-related
QTc prolongation at both therapeutic and some￾what supratherapeutic plasma concentrations. Mean
QTcF was between 10 and 15 milliseconds on
day 4 and between 18 and 25 milliseconds on day
6, and the concentration-QTc relationship was posi￾tive and statistically significant, 0.018 milliseconds per
ng/mL. The prolongation was more pronounced in
the latter part (Tpeak-Tend) of the T wave, and the
JTpeak interval was only mildly affected, with the
largest mean JTpeak_c across all times on both
days only 8 milliseconds and the concentration-JTpeak
relationship very shallow, slightly negative, and not
statistically significant. Our findings thereby confirm
that equipotent calcium and hERG inhibition may
result in QTc prolongation without much of an effect
on the first part of the T wave, the JTpeak interval.
This electrocardiographic pattern has been suggested to
reduce or eliminate the proarrhythmic potential of pure
hERG inhibition,19 which would be consistent with
published clinical experience with perhexiline. However,
it should be acknowledged that mild QTc prolongation
without JTpeak prolongation can also be observed with
hERG-blocking drugs that also affect the peak sodium
current, seen as QRS widening.25 The discrepancy of
the findings in the in vivo assay in human cardiomy￾ocytes in which QTc shortening was observed with the
results from the TQT study is difficult to understand,
and it can be speculated that the distribution of cardiac
ion channels was altered by the harvesting procedure,
compared with the beating heart. In this context, it can
be noted that the current draft ICH S7B/E14 guidance
document recommends standardized ion channel as￾says and in vivo animal studies as primary nonclinical
assays, with the potential to use in vitro repolarization
assays in follow-up studies.26
The clinical experience with perhexiline comes from
decades of use in Australia and New Zealand in pa￾tients with ischemic heart disease or HCM, that is, in
populations at increased risk of malignant ventricu￾lar arrhythmias.4 The Australian regulatory authority,
Therapeutic Goods Administration, collects data on
adverse events associated with prescribed drugs. Al￾though these data are subject to the usual problems as￾sociated with a voluntary reporting system, it provides
some potentially useful information in the analysis
of perhexiline’s safety. The overwhelming majority of
perhexiline toxicity occurs as a result of accumulation
in patients who are genotypic poor metabolizers of
CYP2D6, and therefore, the regular measurement of
plasma levels is recommended. If poor metabolizers
are excluded from the clinical use of perhexiline, the
attainment of supratherapeutic plasma levels is less
likely. A total of 143 adverse drug reports for perhex￾iline were filed between 1975 and 2011, none of which
were reports of arrhythmias associated with its use.
However, n this context, it is important to recognize the
challenges and limitations associated with the reporting
of TdP, especially in a population with a high risk of
polymorphic ventricular tachycardia and sudden death
associated with coronary ischemia.27,28 The Australian
package insert notes QT prolongation as a risk, but
the level of this effect has not been well characterized.
This is apparently based on reports of QT prolongation
with perhexiline and a single case report of TdP asso￾ciated with perhexiline use.8 In this case, a 69-year-old
man with a history of aortic valve replacement and 2
prior coronary artery bypass graft operations presented
with a 2-hour history of chest pain. He had been on
maintenance perhexiline at therapeutic concentrations
and a QTcF of 490 milliseconds. During the initial 24
hours of hospitalization, QTcF increased to 540 mil￾liseconds, and episodes of TdP were observed that were
abolished with overdrive pacing. Discontinuation of
perhexiline was associated with no further episodes of
torsades de pointes. In contrast, there are also reports
on reduction of malignant ventricular arrhythmias with
perhexiline in patients with ischemic heart disease.11 In
4 separate publications during the 1970s, a reduction in
ventricular arrhythmias associated with ischemic heart
disease, mainly premature ventricular contractions, was
reported.9,10,29,30 Thus, despite convincing evidence of
hERG channel inhibition at clinically relevant concen￾trations, TdP has been reported in only a single patient
treated with perhexiline in decades of clinical usage.30
In WHO’s database VigiAccess,31 which lists suspected
adverse drug reactions collected by national drug au￾thorities in more than 110 participating countries, as
of June 15, 2021, there have only been 4 cases of
cardiac arrest and 1 of ventricular arrhythmia, none
with the specific terms of TdP or QT prolongation .
In territories where perhexiline is available, the risk of
proarrhythmia appears to be low in patients who may
benefit from perhexiline therapy, provided that plasma
concentrations are monitored.
Although a TQT study remains the best clinical
predictor of a drug’s effect on the QT interval, its
ability to predict the potential for a proarrhythmia of
a mixed cardiac ion channel blocker is less certain. It
is therefore possible that the effects of hERG blockade
may be mitigated by the inhibitory effects of perhexiline
at another cardiac ion channel.
Conclusions
Perhexiline interacts with multiple cardiac ion chan￾nels resulting in a net lengthening of the QTcF in￾terval clinically. In the TQT study, perhexiline caused
Midei et al 11
prolongation of the QTc interval through lengthening
of the latter part of the interval, the Tpeak-Tend subin￾terval, whereas the JTpeak subinterval was shortened.
This pattern has been suggested to be indicative of
mixed ion channel inhibition and to reduce the proar￾rhythmic risk of mildly QTc prolonging drugs, which
would be consistent with decades of clinical experience
with perhexiline.
Conflicts of Interest
At the time of the work, M.M., G.A., and W.D. were
employees of Heart Metabolics. B.D. is a consultant for and
owns stock and is eligible for stock options in ERT. R.B. and
J.P.C. are employees of ERT. P.S. and J.C. were consultants
for Heart Metabolics.
Funding
The study was sponsored by Heart Metabolics USA, Inc., San
Francisco, USA. No other funding was received.
Data Access
Access to data from this study may be raised with the
corresponding author.
References
1. Horowitz JD, Chirkov YY. Perhexiline and hypertrophic car￾diomyopathy: a new horizon for metabolic modulation. Circula￾tion. 2010;122(16):1547-1549.
2. Phan TT, Shivu GN, Choudhury A, et al. Multi-centre ex￾perience on the use of perhexiline in chronic heart failure
and refractory angina: old drug, new hope. Eur J Heart Fail.
2009;11(9):881-886.
3. Lee L, Campbell R, Scheuermann-Freestone M, et al. Metabolic
modulation with perhexiline in chronic heart failure: a random￾ized, controlled trial of short-term use of a novel treatment.
Circulation. 2005;112(21):3280-3288.
4. Ashrafian H, Horowitz JD, Frenneaux MP. Perhexiline Cardio￾vasc Drug Rev. 2007;25(1):76-97.
5. Grima M, Velly J, Decker N, Marciniak G, Schwartz J. In￾hibitory effects of some cyclohexylaralkylamines related to
perhexiline on sodium influx, binding of [3H]batrachotoxinin
A 20-alpha-benzoate and [3H]nitrendipine and on guinea
pig left atria contractions. Eur J Pharmacol. 1988;147(2):173-
185.
6. Barry WH, Horowitz JD, Smith TW. Comparison of negative
inotropic potency, reversibility, and effects on calcium influx of
six calcium channel antagonists in cultured myocardial cells. Br
J Pharmacol. 1985;85(1):51-59.
7. Walker BD, Valenzuela SM, Singleton CB, et al. Inhibition
of HERG channels stably expressed in a mammalian cell line
by the antianginal agent perhexiline maleate. Br J Pharmacol.
1999;127(1):243-251.
8. Kerr GD, Ingham G. Torsade de pointes associated with per￾hexiline maleate therapy. Aust N Z J Med. 1990;20(6):818-820.
9. Pickering TG, Goulding L. Supression of ventricular extrasys￾toles by perhexiline. Br Heart J. 1978;40(8):851-855.
10. Myburgh DP, Goldman AP. The anti-arrhythmic efficacy of
perhexiline maleate, disopyramide and mexiletine in ventricular
ectopic activity. S Afr Med J. 1978;54(25):1053-1055.
11. Ten Eick RE, Singer DH. Proceedings: Effects of perhexiline
on the electrophysiologic activity of mammalian heart. Postgrad
Med J. 1973;49(Suppl 3):32-43.
12. ICH E14: The Clinical Evaluation of QT/QTc Interval
Prolongation and Proarrhythmic Potential for Non￾Antiarrhythmic Drugs. http://www.ich.org/fileadmin/
Public_Web_Site/ICH_Products/Guidelines/Efficacy/E14/
E14_Guideline.pdf. Published 2005. Accessed June 16, 2021.
13. ICH E14 Questions & Answers (R3) December 10, 2015.

http://www.ich.org/fileadmin/Public_Web_Site/ICH_Products/

Guidelines/Efficacy/E14/E14_Q_As_R3__Step4.pdf. Accessed
June 16, 2021.
14. Hoch M, Darpo B, Brossard P, Zhou M, Stoltz R, Dingemanse
J. Effect of ponesimod, a selective S1P1 receptor modulator,
on the QT interval in healthy individuals. Basic Clin Pharmacol
Toxicol. 2015;116(5):429-437.
15. Guldenring D, Finlay DD, Strauss DG, et al. Transforma￾tion of the Mason-Likar 12-lead electrocardiogram to the
Frank vectorcardiogram. Conf Proc IEEE Eng Med Biol Soc.
2012;2012:677-680.
16. Couderc JP, Ma S, Page A, et al. An evaluation of mul￾tiple algorithms for the measurement of the heart rate
corrected JTpeak interval. J Electrocardiol. 2017;50(6):769-
775.
17. Johannesen L, Vicente J, Gray RA, et al. Improving the assess￾ment of heart toxicity for all new drugs through translational
regulatory science. Clin Pharmacol Ther. 2014;95(5):501-508.
18. Hochberg Y, Benjamini Y. More powerful procedures for
multiple significance testing. Stat Med. 1990;9(7):811-818.
19. Vicente J, Hosseini M, Johannesen L, Strauss DG. Electrocar￾diographic biomarkers to confirm drug’s electrophysiological
effects used for proarrhythmic risk prediction under CiPA.
J Electrocardiol. 2017;50(6):808-813.
20. Vicente J, Zusterzeel R, Johannesen L, et al. Mechanistic model￾informed proarrhythmic risk assessment of drugs: review of the
“CiPA” initiative and design of a prospective clinical validation
study. Clin Pharmacol Ther. 2018;103(1):54-66.
21. Vicente J, Zusterzeel R, Johannesen L, et al. Assessment of
multi-ion channel block in a phase-1 randomized study de￾sign:results of the CiPA phase 1 ECG biomarker validation
study. Clin Pharmacol Ther. 2018;105(4):943-953.
22. 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(3):
369-379.
23. Vicente J, Strauss DG, Upreti VV, Fossler MJ, Sager PT,
Noveck R. The potential role of the J-Tpeak interval in proar￾rhythmic cardiac safety: current state of the science from the
American College of Clinical Pharmacology and the Cardiac
Safety Research Consortium. J Clin Pharmacol. 2019;59(7):909-
914.
24. Fish FA, Gillette PC, Benson DW, Jr. Proarrhythmia, cardiac
arrest and death in young patients receiving encainide and
flecainide. The Pediatric Electrophysiology Group. J Am Coll
Cardiol. 1991;18(2):356-365.
25. Darpo B, Benson C, Brown R, et al. Evaluation of the effect
of 5 QT-positive drugs on the JTpeak interval—an analysis of
ECGs from the IQ-CSRC study. J Clin Pharmacol. 2020;60(1):
125-139.
26. ICH E14/S7B Implementation Working Group. Clinical and
Nonclinical Evaluation of QT/QTc Interval Prolongation and
Proarrhythmic Potential Questions and Answers. Draft ver￾sion, August 2020. https://database.ich.org/sites/default/files/
ICH_E14-S7B_QAs_Step2_2020_0827_0.pdf. Accessed June
16, 2021.
12 The Journal of Clinical Pharmacology / Vol 0 No 0 2021
27. Darpo B. Detection and reporting of drug-induced proarrhyth￾mias: room for improvement. Europace. 2007;9(4):iv23-iv36.
28. Sager PT. Key clinical considerations for demonstrating the
utility of preclinical models to predict clinical drug-induced
torsades de pointes. Br J Pharmacol. 2008;154(7):1544-1549.
29. Drake FT, Haring O, Singer DH, Dirnberger G. Proceedings:
Evaluation of anti-arrhythmic efficacy of perhexiline maleate
in ambulatory patients by Holter monitoring. Postgrad Med J.
1973;49(Suppl 3):52-63.
30. Sukerman M. Proceedings: Clinical evaluation of perhexi￾line maleate in the treatment of chronic cardiac arrhythmias
of patients with coronary heart disease. Postgrad Med J.
1973;49(Suppl 3):46-52.
31. World Health Organization UMC. Perhexiline: Reported Sus￾pected Adverse Drug Reactions. http://www.vigiaccess.org. Ac￾cessed June 16, 2021.
Supplemental Information
Additional supplemental information can be found by click￾ing the Supplements link in the PDF toolbar or the Supple￾mental Information section at the end of web-based version
of this article.