Genetics of QT Response to Moxifloxacin

I. Background and Significance A. Historical Background and Scientific Basis Background The
most common cause of withdrawal or restriction of drugs that are already on the market is
prolongation of the QT interval, and the consequent, potentially fatal, arrhythmia, torsade
de pointes1. First described in the 1960s with quinidine therapy2, torsade de pointes occurs
most commonly with antiarrhythmics3, although sudden cardiac death risk is increased 270%
with use of a non-cardiac QT-prolonging medication4.

The QT interval of the electrocardiogram reflects ventricular myocardial repolarization, and
abnormalities in the duration of the QT interval are the most common indicator of abnormal
repolarization. Although a number of hypotheses exist for how abnormal repolarization, and
specifically QT prolongation, results in ventricular arrhythmias, the most consistent finding
appears to be that during prolonged depolarization sodium channels recover from inactivation
and reactivate, causing what are termed early afterdepolarizations. When combined with
heterogeneity in repolarization as well, these early afterdepolarizations create a favorable
myocardial substrate for reentry, resulting in propagation of intramyocardial reentry waves
and torsade de pointes5.

On a cellular level, the ion channels most directly associated with repolarization are
potassium-conducting, and thus most medications and genes associated with QT prolongation
have effects on potassium currents. Specifically, blockade of the HERG/KCNH2 (IKr) ion
channel has been implicated in the majority of drug-induced QT prolongation6, although other
ion currents are involved in congenital long QT syndrome, including the IKs (KCNQ1 and
KCNE1), INa (SCN5A), and IK1 (KCNJ2).

Sudden cardiac death as a result of QT interval prolongation and subsequent ventricular
arrhythmia (torsade de pointes) is a devastating adverse effect of many common medications7,
8. Drug-induced QT prolongation is the number one barrier for new therapeutic agents making
it to market. While a number of medications, environmental factors, and genetic factors have
been associated with QT prolongation, our ability to predict on an individual basis who will
develop QT prolongation, not to mention torsade de pointes, is limited.

Human genetics has provided an opportunity to change this paradigm by enabling the discovery
of individuals at risk for toxicity. The continuous QT interval is heritable9, with multiple
environmental and genetic contributors. We have used genome-wide association studies (GWAS)
to identify over 60 common polymorphisms that collectively explain more of QT interval than
all known clinical factors combined. However, our knowledge of the application of this
information to the patient level is incomplete.

A personalized genetic approach to toxicity prevention would be important because cardiotoxic
drug response is not specific to a single class of drugs; risk prediction for
arrhythmogenicity can be applied across medications used in a variety of conditions.
Narrowly, this research can enable drugs with currently marginal risk/benefit profiles to be
brought to market, sparing those at-risk and providing access to new therapies for those who
are not. From a public health perspective, it will reduce potentially fatal toxicity and thus
improve human health through identification of particular at-risk individuals. But more
broadly the application of human genetics requires rigorous definitions of association and
well-powered and -designed tests of pharmacogenetics before broad application can be
considered.

B. Previous Studies In a meta-analysis of three prospective cohorts--Cardiovascular Health
Study, the Framingham Heart Study, and the Rotterdam Study--in 13,685 white people of
European descent as part of the QTGEN consortium, Newton-Cheh and colleagues identified
common variant associations (p < 5x10-8) at five loci previously associated with QT interval:
NOS1AP, KCNQ1, KCNE1, KCNH2, and SCN5A, as well as new associations in 5 other loci
previously unrecognized to influence myocardial repolarization10. At these 10 loci, 14
independent variants explained 5.4 - 6.5% of the variation of the QT interval, which was more
than is explained by sex or age, the strongest non-genetic clinical factors10. A QT genotype
score based on these variants was associated with a 9.7 - 12.4ms longer QTc in the top
quintile compared with the bottom quintile in the meta-analysis samples, which included
heterogeneous age, risk factor and drug exposure profiles10-12. These effects were consistent
in individuals of both European and African American ancestry. This QT score was
independently validated in a Finnish population sample in which complete medication
ascertainment enabled excluding those on QT-altering therapy, and in which the QT score was
associated with a 15.6ms QT interval difference between the top and bottom quintiles13. In
this analysis, the correlation of the measured effect estimates to those of the original
association study from which they were obtained was 0.99. The QT based on the genotype score
was a significant predictor of the actual QT interval measured (P < 10-107). This study is
important for the current proposal because a) it demonstrates the consistency of genetic
effect estimates derived from meta-analysis of heterogeneous studies when applied to
independent samples and b) it confirms that excluding individuals with QT-altering drug
therapy can reduce the noise from non-genetic sources of QT variation and improve the genetic
signal and thereby increase power.

Moxifloxacin is well-suited to study drug-induced QT prolongation, as it is known to cause
transient and mild QT prolongation. Compared to placebo, 400mg of oral moxifloxacin is
associated with an approximately 10 msec increase in the heart rate-corrected QT interval14,
15. Despite this increase in QT interval, there is no reported increased risk of sudden death
with oral administration of a single dose of moxifloxacin16, so it is a safe and
well-validated drug to use in the study of cardiotoxic drug response in healthy human
subjects. Moreover, it has been widely used by the pharmaceutical industry as a positive
control as required by the FDA to demonstrate a sufficiently sensitive method to detect QT
prolongation.

We have demonstrated QT interval prolongation to moxifloxacin in a pilot study of 20 subjects
recruited at Massachusetts General Hospital. Apparently healthy male and female volunteers
between 18 and 50 years old were eligible if they: were free of known cardiovascular, renal,
hepatic disease, had no personal or family history of sudden cardiac death, used no
prescribed or over-the-counter medications, had no bradycardia or QTc prolongation on
electrocardiography, and had a normal potassium and magnesium. Subjects received 400 mg of
oral moxifloxacin or placebo on different days at exactly 9AM in the morning (diurnal
variation in QT interval is well recognized17) in the fasting state (to avoid interference
with absorption) and had six ten-second electrocardiograms after a minimum of ten minutes of
rest, recorded every half hour for six hours. Mass spectrometric analysis of moxifloxacin
levels in plasma at 2 and 6 hours after administration closely matched prior reports18 with
tight ranges consistent with its known high bioavailability (2hr: 2.8 (SD 0.47) and 6hr 2.64
(SD 0.48) ng/μL). We observed a 12.3 msec increase in heart rate-adjusted QT interval
averaged over hours 2-6 compared to baseline after moxifloxacin exposure compared to placebo
exposure (ΔΔQTc = +12.3 msec, SD 8.3; Noseworthy, manuscript in preparation).

C. Rationale of Proposed Benefit of Research Drug-induced QT prolongation is a significant
public health risk and a major barrier to drug development19-21. For example, cisapride, an
esophageal motility agent used for gastro-esophageal reflux, may have resulted in more than
80 deaths before it was pulled from the market22. In addition to identification of culprit
medications, there are also characteristics of the vulnerable patient, including female
sex23, bradycardia24, hypokalemia25, and genetic predisposition by rare mutations3, 6, 26,
that place certain individuals at higher risk of fatal arrhythmia. Identification of these
at-risk individuals with the use of genetic markers would represent a critical advance toward
safer pharmacotherapy7. The basis of this study is to expand the use of genetics beyond rare
family mutations that predispose to risk of QT prolongation to large-scale screening of the
general population by genotyping (assaying for specific variants) for relatively common
polymorphisms that could increase risk of QT prolongation, and subsequent torsade de pointes.

II. Specific Aim We hypothesize that common genetic variants with intermediate effects on
resting QT, when examined in aggregate, can identify a subgroup of individuals at risk of
exaggerated prolongation of the QT interval in response to a QT-prolonging medication.

This hypothesis will be tested through the following specific aim:

To assess the ability of common genetic variants in aggregate to predict drug-induced QT
prolongation in healthy subjects. We will recruit 80 healthy volunteers drawn from the top
and bottom quintiles of a QT genotype score for assessment of QT response to moxifloxacin
compared to placebo at the Massachusetts General Hospital (MGH).

III. Subject Selection A. Inclusion/Exclusion Criteria: Through a separate protocol, Dr.
Newton-Cheh has recruited to date over 1000 healthy volunteers aged 18 to 40 with consent for
re-contact on the basis of genotypes determined after an initial screening visit, all of whom
have undergone DNA extraction. From the existing collection of 1000 subjects, approximately
75% are self-described European/Caucasian and 8% are African Americans. These are otherwise
healthy individuals ages 18 and older. For this study, we will genotype 68 independent QT
SNPs using the Sequenom genotyping array in the MGH Center for Human Genetic Research
Sequenom core on these samples. The genotype score is the sum of the predicted effects on QT
interval for each genotype. For example, SNP rs12143842 is a C/T SNP; for each copy of the T
allele the QT interval is 3.50 msec higher such that individuals with the TT genotype have a
QT interval 7 msec higher than those with CC genotype and individuals with the CT genotype
have a QT interval 3.5 msec higher. These predicted effects are then determined for all 68
SNPs and summed to create a single number, the predicted change in QT interval on the basis
of those genetic variants. From the genotype score, we will determine the set of individuals
who belong to the top or bottom quintile of QT genotype risk score to be eligible for our
moxifloxacin study (see below for details of the genotype score calculation). Apparently
healthy male and female volunteers of European and African American ancestry, between 18 and
50 years old, will be eligible if they are free of known cardiovascular, renal, hepatic
disease, have no personal or family history of sudden cardiac death; use no prescribed or
over-the-counter medications as well as recreational drugs; have no bradycardia (defined as
resting heart rate < 50 bpm), conduction disease (QRS > 100ms) or QTc prolongation on
electrocardiography (QTc > 500msec); and have a normal serum potassium (K>3.3) and magnesium
(Mg>1.8), as well as renal and liver function tests. We will exclude woman who are nursing,
pregnant or planning to become pregnant during the study period, counsel them on the
importance of using birth control during the study period, and check a serum HCG on the
screening visit. We estimate that 266 individuals (133 in the highest and lowest quintiles,
respectively) will be eligible.

IV. Subject Enrollment

A. Methods of Enrollment and Procedures for Informed Consent:

See above for details of subject identification for enrollment. Eligible individuals will be
invited to come in for a screening visit for the moxifloxacin study with a target enrollment
of 40 in each genotype group. We plan to contact up to 266 individuals who are in the upper
and lower quintiles of QT genotype score to come in for the screening visit. We anticipate
inviting approximately 160 individuals for the screening visit, with a goal enrollment sample
size of 80 individuals. These subjects will be contacted by phone about participation in this
study, and then informed consent will be obtained on arrival for the screening visit, and
verified at subsequent visits if they are eligible for the study protocol.

V. Study Protocol A. Data Collection: Subjects will be brought to the MGH Clinical Research
Center for an initial screening visit in which eligibility criteria will be reviewed (with
exclusion/inclusion as applicable), and they will be given an ECG recording. Eligible
individuals for the study visits will be brought back later to the MGH Clinical Research
Center on two separate days separated by at least one week (to ensure washout). On each
visit, they will receive study drug (either 400 mg oral moxifloxacin or placebo) allocated by
double blinded block randomization (by sex and genotype group separately) at 10AM in the
fasting state (statistician will hold the randomization key; all other study personnel will
be blinded until study closure). Women will be screened for pregnancy using a urine HCG at
the time of both study visits. They will undergo 6 ECGs at time 0 and at 30 min intervals
thereafter, in the seated position after at least 15 min rest using fixed electrode placement
for a total of 6 hours following administration of study drug. At 2, 4 and 6 hours we will
draw a blood sample for plasma moxifloxacin determination by mass spectrometry. All ECGs are
uploaded to a research partition of the GE MUSE 7 ECG database of the Massachusetts General
Hospital, and the 12 SL algorithm applied for interval measurement, as used in prior studies
by our group. We will monitor the QT every 30 minutes throughout the protocol and will
identify any subjects with marked QT prolongation. Any subject who has a QTc >500 ms will be
kept in the HCRC for further observation, with the decision to either follow-up at the HCRC
every morning until it is less than 500ms or be admitted to the hospital for telemetry based
on the degree of prolongation and clinician's judgment.

B. Venipuncture: Venipuncture will be performed using standard techniques to obtain plasma
for quantification of moxifloxacin levels. We will collect 10 mL of blood at 2hrs, 4hr and
6hrs after study drug administration (Total 30mL).

C. Genotyping: Genotyping will be performed by the Sequenom genotyping array in the MGH
Center for Human Genetic Research Sequenom core on existing genomic DNA isolated in.

VI. Biostatistical Analysis:

QT Genotype Score. We plan to use a modification of the QT genotype score that has been
previously established11, 12 and validated by us in a separate Finnish cohort13. The score is
constructed using allele copy number and effect estimates using the following formula (Table
1, next page):

QTscore = [(SNP1 allele copy number)*(SNP1 effect estimate in ms)] + [(SNP2 allele copy
number)*(SNP2 allele effect estimate in ms)] + … [(SNP68 allele copy number)*(SNP68 allele
effect estimate in ms)] From the genotype score, we will determine the set of individuals who
belong to the top or bottom quintile of QT genotype risk score (of whom, we expect 75% (n=133
per quintile)) to be eligible for our moxifloxacin study, based on our prior pilot work. We
will invite these individuals to come in for a screening visit (see above) for the
moxifloxacin study with a target enrollment of 40 in each genotype group.

Outcome assessment. Any individual with sinus arrhythmia present in over 50% of ECGs during
the two study visits will be dropped from analysis. ECGs with premature ventricular or atrial
beats will be dropped from analysis (these are very uncommon in healthy volunteers). The QTc
using Fridericia's heart rate correction (QTc = QT/3√RR) will be taken as the average of all
eligible ECGs from each time point. The ΔQTc will be calculated as the difference in QTc from
baseline (after study drug/placebo administration) for each time point. From the ΔQTc, the
difference between the post-moxifloxacin and post-placebo ΔQTc will then be calculated at
each time point (ΔΔQTc) as well. The mean ΔΔQTc from the 180-300 minute time points (which
from our pilot work and published reports is expected to be 10 msec) will then be compared
between the two genotype groups (top and bottom and quintile) using an unpaired t-test. We
estimate that 40 people in each group will be needed to have adequate power to detect a
clinically meaningful (6.3 msec) difference in QT-prolongation between the two groups based
on our pilot data (Table 2). An increase of 6 msec after exposure to a drug compared to
placebo is the threshold at which the FDA raises concern for the potential of a drug to cause
torsade de pointes and is thus clinically significant.

Secondary analyses. We plan to perform three secondary analyses. First, we will examine the
influence of single SNPs of stronger effect on QT response to moxifloxacin. As power
calculations, and study enrollment are based on the aggregated genotype score, we expect
these secondary analyses to be underpowered. Second, if we observe a significant difference
in QT response between top and bottom quintiles of QT genotype score, we will examine the
influence of deciles on observed results comparing the 9th (80-90%ile) and 10th deciles
(>90%ile) of score to the bottom decile or quintile, although given the composite nature of
the genotype score and additive nature of the variants, we do not expect a threshold effect.
Lastly, we will determine the impact of baseline QTc on QT response and test whether
additional adjustment for baseline QTc alters any observed effect of genotype score on QT
response to moxifloxacin.

Anticipated results. Successful completion of the primary analysis of Specific Aim 1 would
identify that individuals in the top quintile of QT genotype score demonstrate greater QT
prolongation in response to exposure to moxifloxacin than those in the bottom quintile. This
would have important implications for the use of genetic predictors in understanding and
management of drug-induced QT prolongation. For one, it would demonstrate in principle that a
simple genetic test is predictive of risk of drug-induced QT prolongation. Such an analysis
could be performed on patients prior to the use of currently approved QT-prolonging
medications. Second, it would demonstrate that genes associated with QT prolongation at
baseline are also associated with risk of drug-induced QT prolongation. This finding would
have implications at both a risk-predictive level, as above, but also in our physiological
understanding of the mechanisms of drug-induced QT prolongation.

Inclusion Criteria:

- Healthy volunteers

- Genotype in the highest and lowest quintiles of genetic predictors of QT interval
duration

- Able to swallow pills

Exclusion Criteria:

- Inability to provide informed consent

- Prior known cardiovascular, renal, hepatic disease

- Personal or family history of sudden cardiac death

- Current use of prescribed or over-the-counter medications as well as recreational
drugs

- Resting bradycardia (defined as resting heart rate < 50 bpm)

- Conduction disease (QRS > 100ms)

- QTc prolongation on electrocardiography (QTc > 500msec)

- Abnormal potassium or magnesium serum level

- Abnormal renal or liver function tests

- Women who are nursing, pregnant or planning to become pregnant during the study period

- Tendon disorder or rupture