Unresolved procedural, regulatory and statistical issues in assessing human QT prolongation and performing the ‘thorough

1. Unresolved procedural, regulatory and statistical issues in assessing human QT prolongation and performing the ‘thorough QT study’ Borje Darpo MD PhD, FESC Associate Professor in Cardiology Pharmaceutical Consultant borje.darpo@telia.com

2. Objective of presentation • To outline some areas within clinical QT assessment, which I believe need more studies and publicly shared data to allow for definitive recommendations‘More than one road lead to Rome • – • until someone builds a motorway’ • Hieronymus ‘Filibuster’ Hippocampus A.D. -07

3. Some unresolved issues – not necessarily the most important ones Thorough QT study Baseline assessment Heart rate correction algorithm for drugs without or with a small inherent effect on the heart rate The role of the positive control Which effect of the positive control establishes assay sensitivity? How should emerging new techniques be validated?

4. 1. Baseline in the TQT study Background: Adjustment for baseline measurements is important for reducing the influence of inter-subject differences and general, as well as study-specific, diurnal effects such as those due to food Currently ‘recommended’(code for ‘recently requested by the FDA’s IRT’):Cross-over studies:Predose baseline immediately before dosing on treatment day in each period Parallel studies:One full, ‘time-matched’ baseline on the day before dosing in each treatment group

5. Baseline in parallel TQT studies Pre-dose baseline:Immediately before dosing on Day 1 in each Tx group. Time-matched baseline (‘recommended’):Recorded at same time-points on Day -1 as after dosing on Day 1. Time-averaged: Recorded as above, but baseline derived from average of Day -1 values

6. Baseline in parallel TQT studies In manuscript Darpo, Ferber, Sarapa

7. Baseline adjusted, placebo corrected QTcF after 400 mg moxifloxacin for 7 days SD varied between 11 and 13 across baselines and Tx

8. Ratio between SD’s predose or averaged / time-matched Baseline assessment in parallel TQT studies Median SD values ranged from 8 to 17 across studies and Tx Conclusion: Variability consistently lower with averaged baseline compared to time-matched or predose

9. Baseline in parallel TQT studies Conclusions: ‘Averaged’ baseline reduced total variability more than time-matched, thus allowing for more efficient study designs If the subject-specific part of circadian variability was substantial, time-matched BL would have significantly reduced the total variability Our data suggest that the subject-specific circadian variability was relatively small and ‘averaging’ therefore resulted in an over-all lower variability

10. 2. Heart rate correction algorithm for drugs without an effect on the heart rate • Full time-matched baseline is often advocated in cross-over studies to allow for individual heart rate correction (QTcI) • often also recommended for drugs without effect on the heart rate • ECGs are often recorded after 10-20 minutes of supine rest • which generates relatively narrow heart rates recorded at rest This approach generates QTcI values, • which rarely are different from QTcF for drugs without or with a small effect on the heart rate, but certainly requires more ECGs and 1 additional study day/Tx period • for which the utility for drugs with an effect on the heart rate can be questioned (or at least warrants further studies). • There is yet no general agreement on how to study the QTc effect with drugs with an inherent effect on the heart rate, through e.g. autonomic alteration

11. 3. Why are we using a positive control? According to ICH E14: “The confidence in the ability of the study to detect QT/QTc prolongation can be greatly enhanced by the use of a concurrent positive control group to establish assay sensitivity“ “…, the positive control (whether pharmacological or non- pharmacological) should be well-characterized and consistently produce an effect corresponding to the largest change in the QT/QTc interval that is currently viewed as clinically not important to detect (a mean change of around 5 ms or less)”

12. Establishing assay sensitivity Uncontroversial in X-over studies with short duration in which + control can be included as separate Tx period Challenging in studies with long duration of Tx or long wash-out (e.g. dose escalation, drug accumulation) Recently the IRT has requested running the positive control in separate Tx group in parallel TQT studies which means that 67% of subjects are exposed to either placebo or one dose of moxifloxacin

13. IRT request for parallel-group studies Drug Baseline Day 21 Subjects 1 - 40 Placebo Baseline Day 21 Subjects 41 - 80 67% Separate + control Baseline Day 21 Subjects 81 - 120 Day of primary assessment

14. Question: • Does the gain with separate + control Tx-group justify the added cost and complexity of the study? • Could an alternative approach be acceptable? • All Tx periods are extended by 1 day and a single-dose of • + control/placebo is added on this day • Issue: Baseline adjustment and placebo-correction will be within group for + control effect but not for drug effect

15. Parallel-group studies Drug Baseline Day 21 Subjects 1 - 40 Placebo/+ control Baseline Day 21 Subjects 41 - 80 X Day of primary assessment Means 33% less subjects

16. 4. How does the + control establish assay sensitivity? • FDA IRT: The effect should be comparable to other similar studies using moxifloxacin and an effect > 5 ms should be demonstrated (lower CI > 5 ms). • Health Canada: The peak effect of the + control should be ‘around 5 ms’ and the lower bound of the CI above 0 ms • Others 1: There is no difference between a peak effect and an effect at other time-points. An statistically significant effect ‘around’ 5 ms at any time point is therefore sufficient • Others 2: The precision of the effect should allow detection as small as 5 ms, i.e. the width of the CI should be < + 5 ms.

17. 5. How to validate emerging techniques? Submitted 2008. Fosser, Duczinski, Agin, Wicker, Darpo Different methods generate different absolute values

18. 5. How to validate emerging techniques? Apply to data set from more than one TQT study and compare the standard output (time-matched, BL adjusted, placebo-corrected QTcF) with accepted method Result:Good agreement(can be quantitativelydefined) Submitted 2008. Fosser, Duczinski, Agin, Wicker, Darpo

19. Difference from Placebo in QTcF: Study 5 25 Method 1 Method 2 20 15 QTcF Difference (ms) 10 5 0 0 2 4 6 8 10 12 14 Time Post Dose (hours) Another example…. Not so good agreement Cross-over, 4 days of moxifloxacin 400 mg qd

22. How to validate new techniques? Recommendations Validate new measurement methods (techniques) on data set from TQT studies Ideally, same dataset(s) for all…. Look at data several ways, e.g. Absolute values in drug-free condition Time-matched QTcF Bland Altman plots (slope, LoA) Proportion of categorical outliers Create quantitative standards for output

23. Conclusions Many of these topics can be studied, thereby allowing unbiased recommendations; The generation of a publicly accessible database comprised of several moxifloxacin / placebo datasets from TQT studies will greatly facilitate this research; Much of this research is ‘precompetitiv’ research – all findings rapidly become public and will not provide anyone individual player competitive edge; Sponsors with experience from many TQT studies can obviously also contribute individually and are encouraged to publish their results.

24. THANK YOUBorje Darpo MD PhDAssociate Professor of Cardiology • Pharmaceutical Consultant • borje.darpo@telia.com