Thorough QT Study: Design Features for Consideration - IIR

Slide 1
Thorough QT Study:
Design Features for Consideration
Charles M. Beasley, Jr., M.D.
Distinguished Lilly Scholar
Chief Scientific Office, Global Product Safety
Eli Lilly and Company

Slide 2
Ventricular Repolarization: Difficulties in Biomaker
Measurement and Analysis
• “there are difficulties in the exact determination of the points
which are to be used for the measurement of the QT interval
in a given complex”
• “there are difficulties when the actual QT duration is
corrected for heart rate”
Lepeschkin & Scuawicz (1952) Circulation 6: 378-388

Slide 3
Ventricular Repolarization: Difficulties in Biomaker
Measurement and Analysis (Continued)
• The range of complex-to-complex variability in QT
measurement may be 25 msec.
Malik & Camm (2001) Drug Safety 24:323-351
BUT
• The difference of interest between mean QTc values is as
small as 5 msec.
Malik (2001) J Cardiovasc Electrophysiol 12:411-420

Slide 4
Features to Consider
• Desired outcome
• Homogeneous vs diverse subjects
• Parallel vs Cross-over design
• Number of replicate ECGs signal averaged
• Method of correcting QT interval for heart rate
• Which positive control
• Complete blinding vs blinding of ECG readers
• Can adequate exposure to test drug be achieved with a single dose or does the test drug
require multiple administrations to achieve steady state
• What value is compared between treatments
• Time points of data acquisition
• Method of protecting against inflated potential for false finding of drug effect inherent in
multiple time point comparisons
• Inclusion of pK assessments

Slide 5
Hierarchy of Desired Outcomes
• “Stages 0-4”
‣ Difficult to conduct a study that will demonstrate a “Stage 0”
outcome
‣ Although “Stages 0-2” would meet ICH-E14 criteria as “negative
(good outcome)” for a thorough QT study, a “Stage 2” outcome
might result in adverse labeling
‣ The closer to “Stage 0”, the better for the commercial
opportunities for a drug intended for wide chronic use in non-life
threatening condition
Morganroth, et al. (2004) Am J Cardiol 93:1378-1383
Beasley, et al. (2005) J Am Coll Cardiol 46:678-687

Slide 6
“Stage 0” Outcome
Difference Drug - Placebo
Upper Bound
1-sided 95% CI
Point Estimate
0
5 10
msec

Slide 7
“Stage 1” Outcome
Difference Drug - Placebo
Upper Bound
1-sided 95% CI
Point Estimate
0
5 10
msec

Slide 8
“Stage 2” Outcome
Difference Drug - Placebo
Upper Bound
1-sided 95% CI
Point Estimate
0
5 10
msec

Slide 9
“Stage 3” Outcome
Difference Drug - Placebo
Upper Bound
1-sided 95% CI
Point Estimate
0
5 10
msec

Slide 10
“Stage 4” Outcome
Difference Drug - Placebo
Upper Bound
1-sided 95% CI
Point Estimate
0
5 10
msec

Slide 11
Homogeneous vs Diverse Subjects
• It might be likely that the more homogeneous the subjects, the less
INTERsubject variability
‣ Genetic screening for long QT syndrome variants?
‣ Use of only very healthy subjects?
• If factors that contribute to INTERsubject variability also have a
differential influence on responses across treatments, they will contribute
to INTRAsubject variability
• It might be likely that using subjects without factors believed to possibly
influence QT length over time, the less INTRAsubject variability
‣ Use of only males with less hormonal variability over time?
• INTRAsubject variability is the primary determinate of the precision of the
measured treatment difference

Slide 12
Parallel vs Cross-over
• More statistical power per subject with cross-over
• Parallel can be run more quickly, requiring less time commitment
from subjects but will require > [(# treatments) * (# subjects)] to
achieve same power as cross-over with (# subjects)
• Cross-over is most frequently used

Number of Replicate ECGs Signal Averaged to Yield an Average
QT Value
Slide 13
•Beat-to-beat variability under optimal recording conditions is
substantial
‣ True variability
‣ Measurement error
• Signal average to strengthen the signal
• How many replicates
‣ 3 is common
‣ Probably some slight increase in reduction of variability with 6-7
‣ The more the better within a time frame where factors influencing QT are not
changing but the increase in reduction of variability will be very small beyond 6-7

Slide 14
Method of Correcting QT for Heart Rate
• Historical population formula
• Experimental population formula
• Individual correction formula
‣ 30+ non-treatment values over range of heart rates
‣ Heart rate range should cover heart rate expected with treatment
• Model based
‣ RR – is a covariate in a repeated measures model
Dmitrienko & Smith (2002) Drug Inf J 36:269-279
Dmitrienko & Smith (2003) Pharm Stat 2:175-190

Slide 15
Which Positive Control
• Produce a mean increase in QTc of 5-10 msec with tight
INTERsubject variability
• Moxifloxacin most commonly used
• IV ibutilide (ultra low dose) can be individually titrated
•Other

Slide 16
Complete Blinding vs Blinding of ECG Readers
• To blind subjects and drug administrators requires many
placebos and/or double-dummy administrations and all treatment
periods must be at equal length
• Generally necessary to completely blind only test drug and
placebo

Slide 17
Can Adequate Exposure to Test Drug be Achieved with a Single
Dose or Does Test Drug Require Multiple Administrations to
Achieve Steady State
• The single dose scenario applies when drug is to be used as
single-dose PRN or is so well tolerated that a large multiple at
intended maximum therapeutic dose can be administered with
impunity
• Impacts the recommended possible primary comparisons

Slide 18
What Value is Compared Between Treatments
• With single dose design, 3 possibilities
• With multi-administration design, 2 possibilities may be more
reasonable
• Change from “baseline” (as opposed to absolute value on
treatment) may not be necessary
‣ “Single-delta” comparison may be adequate

Slide 19
Single Dose Design
• Single-delta
Drug
Placebo
Test Day
Administer
Administer
Time
QTc
V 1
QTc
V 2
V 1
–V 2
• Double-delta
Drug
Placebo
Baseline Day
QTc
QTc
V 3
V 1
Test Day
Administer
Administer
QTc
V 2
(V 2
–V 1
) – (V 4
–V 3
)
QTc
Multiple baseline days may be averaged
Time

Slide 20
Single Dose Design
• Triple-delta
Baseline Day
Test Day
Drug
QTc
QTc
QTc
QTc
V 1
V 2
V 3
Administer
V 4
((V 4
-V 3
) – (V 2
-V 1
)) – (V 8
-V 7
) – (V 6
-V 5
))
Multiple baseline days may be averaged
Placebo
QTc
QTc
QTc
QTc
V 5 V 6
V 7Administer V 8
Time
‣ Using V 3
, V 4
and V 7
, V 8
allow for an alternative double-delta method

Slide 21
Multiple Administrations to Steady State Design
•Single-delta
Drug
Test Day
QTc
Placebo
Administer Administer Administer V 1
QTc
V 1
–V 2
•Double-delta
Administer Administer
Time
Administer V 2
Drug
Placebo
Baseline Day
Test Day
QTc
QTc
V 1
Administer Administer Administer V 2
QTc
QTc
(V2-V1) – (V4-V3)
Multiple baseline days
may be averaged
V 3
Administer Administer Administer V 4
Time
‣Acquisition of QTc immediately before administration of test drug and placebo on day 1 of administration of each
would allow for triple-delta or alternative double-delta but not personally recommended

Slide 22
Time Points of Data Acquisition
• Must reasonably account for individual variation in T max and
possible hysteresis
• The more time points, the greater potential for failing to
demonstrate non-inferiority to placebo at a given time point due to
the bias inherent in multiple comparisons
• 3 time points around population T max and a trough (before a next
administration) to account for metabolites may be reasonable

Slide 23
Protection Against Inflated Potential for False Finding of Drug
Effect in Multiple Comparisons
• Inconsistent with ICH-E14 guidelines
Beasley, et al. (2005) J Am Coll Cardiol 46:678-87
Eaton, et al. (2006) Drug Inf J 40:267-271

Slide 24
Inclusion of pK Assessments
• With each ECG collection
• Allows for pK-pD relationship estimation
‣ May offer some protection against drawing definitive conclusions
from a spurious finding at a given time point among multiple time point
comparisons

Slide 25