1 QT Evaluation Studies: Pharmacometric Considerations Leslie Kenna, Peter Lee and Yaning Wang Office of Clinical Pharmacology and Biopharmaceutics CDER/FDA Clinical Pharmacology Subcommittee Meeting November 17, 2003
2 Outline Overarching question Challenges Methods: Clinical Trial Simulation Preliminary results
3 What do we want to know? Drug effect on QT interval: “Worst-case” scenario
4 Challenges Variation in response > response of interest
5 Challenges Variation in response > response of interest Wide intra-individual variability e.g. #1: within day variability
6 QTcF (msec) Time (hr) Historical Baseline QTc Data for Drug “X” Subject i: 10 ECGs / time 15 msec
7 Challenges Variation in response > response of interest Wide intra-individual variability e.g. #1: within day variability e.g. #2: between day variability
8 QTCF (msec) Time (hr) Smooth Day 1 Smooth Day 4 Smooth Day 3 Smooth Day 2 Smooth through Days 1:4 Intraindividual Variability in Baseline QTc: Subject i: 4 Days of Measurement
9 Challenges Variation in response > response of interest Wide intra-individual variability e.g. #1: within day variability e.g. #2: between day variability Wide inter-individual variability
10 QTCF (msec) Time (hr) Interindividual Variability in Baseline QTc Subject iSubject k
11 Observations in Recent Submissions Diverse study designs: e.g. duration, timing, # replicates
12 “Baseline” for Six QT Evaluation Studies DrugBaseline 1A single Day –14 ECG 2Mean of 3 ECGs at t=0 3Mean of 1 screening ECG & 1 follow-up ECG 4Drug, placebo: Mean of 100 ECGs over 2 baseline days Positive Control: Mean of 10 ECGs 10 min pre-dose 5Mean of 108 ECGs before each treatment arm & 30 ECGs on placebo 6Median of 9 ECGs on placebo
13 Observations in Recent Submissions Diverse study designs: e.g. duration, timing, # replicates Different response to same positive control Case 1Moxifloxacin 400 mg 8 msec QTcF Case 2Moxifloxacin 400 mg13 msec QTcF
14 Observed Response to Moxifloxacin in Two Recent QT Evaluation Studies Case 1Case 2 QTcF at Tmax (95% CI) 8 msec (6,9) 13 msec (9, 17) DemographicsHealthy males Sampling timeCovered tmax moxifloxacin N5945 ECGS / day3: baseline 4: on drug 1: baseline 11: on drug replicates / time63 What role did study design play in the discrepancy in response?
15 Observations in Recent Submissions Diverse study designs: e.g. duration, timing, # replicates Different observed response to same positive control Case 1Moxifloxacin 400 mg 8 msec QTcF Case 2Moxifloxacin 400 mg13 msec QTcF Observed response sensitive to analysis method Mean vs. outlier analysis
16 Mean vs. Outlier Analysis Mean Response QTcF at Tmax vs placebo (90% CI) Drug “X”4 msec (2,5) (+) Control9 msec (8,11)
17 Mean Response QTcF at Tmax vs placebo (90% CI) Outliers # (%) Subjects QTcF > 30 msec Drug “X”4 msec (2,5) 14 subjects (15%) (+) Control9 msec (8,11) 11 subjects (16%) Placebo7 subjects (8%) Mean vs. Outlier Analysis
18 Observations in Recent Submissions Diverse study designs: e.g. duration, timing, # replicates Different observed response to same positive control Case 1Moxifloxacin 400 mg 8 msec QTcF Case 2Moxifloxacin 400 mg13 msec QTcF Observed response sensitive to analysis method Mean vs. outlier analysis Definition of baseline
19 Definition of Baseline Influences Analysis (# Outliers) Treatment – free + On - placebo Placebo 1 1X Dose 2 5X Dose 4 “Baseline” Dose-response appears shallow
20 Definition of Baseline Influences Analysis (# Outliers) Treatment – free + On - placebo Treatment – free Placebo 12 1X Dose 22 5X Dose 48 Response directly proportional to dose “Baseline”
21 Goal Use available data to aid in the prospective design of QT studies
22 Specific Aims Assemble a QT database from submissions Resample the data and use Clinical Trial Simulation to evaluate: (a) clinical trial designs (b) data analysis methods
23 Clinical Trial Simulation Approach
24 Create true data Sample from historical baseline QT data Choose models and parameters for study design, PK, PD Add baseline response to simulated response to treatment Sample true data according to study design Estimate response Metrics in Concept Paper, in submissions Repeat “many” times at given set of study design parameters Compute Performance Power: Fraction of simulations yielding insignificant effect Simulation Study Overview
25 Simulation Study Overview Create true data Sample from historical baseline QT data Choose models and parameters for study design, PK, PD Add baseline response to simulated response to treatment Sample true data according to study design Estimate response Metrics in Concept Paper, in submissions Repeat “many” times at given set of study design parameters Compute Performance Power: Fraction of simulations yielding insignificant effect
26 Simulation Study Overview Create true data Sample from historical baseline QT data Choose models and parameters for study design, PK, PD Add baseline response to simulated response to treatment Sample true data according to study design Estimate response Metrics in Concept Paper, in submissions Repeat “many” times at given set of study design parameters Compute Performance Power: Fraction of simulations yielding insignificant effect
27 Simulation Study Overview Create true data Sample from historical baseline QT data Choose models and parameters for study design, PK, PD Add baseline response to simulated response to treatment Sample true data according to study design Estimate response Metrics in Concept Paper, in submissions Repeat “many” times at given set of study design parameters Compute Performance Power: Fraction of simulations yielding insignificant effect
28 Simulation Study Overview Create true data Sample from historical baseline QT data Choose models and parameters for study design, PK, PD Add baseline response to simulated response to treatment Sample true data according to study design Estimate response Metrics in Concept Paper, in submissions Repeat “many” times at given set of study design parameters Compute Performance Power: Fraction of simulations yielding insignificant effect
29 Simulation Study Overview Create true data Sample from historical baseline QT data Choose models and parameters for study design, PK, PD Add baseline response to simulated response to treatment Sample true data according to study design Estimate response Metrics in Concept Paper, in submissions Repeat “many” times at given set of study design parameters Compute Performance Power: Fraction of simulations yielding insignificant effect
30 Step 1(a): Create True Data (Sample Baseline) 1. Randomly pick a subject from the database 3. Randomly pick n baseline observations / time point 2. Randomly pick a day of subject i’s baseline observations Day 4
31 Step 1 (b): Choose Simulation Conditions 4 treatments evaluated: 2 doses of drug, placebo, active control Parameters to be varied 1. Crossover vs. parallel design 2. Single dose vs. steady state design 3. N 4. ECG sampling Timing # replicate ECGs (n): at baseline, on treatment # days of measurement: at baseline, on treatment 5. PK/PD model for drug Delayed or direct response EC50; parent, metabolite 6. PK model for drug CL; parent, metabolite ka
32 Treatment effect Time QTc Time QTc + + = = Step 1 (c): Simulate Drug Response Baseline QTc Treatment response Drug Time QTc Placebo Time QTc Time QTc Time QTc
33 Step 2: Sample According to Study Design e.g. Sampled Baseline e.g. Sampled Response Time QTc Time QTc Time QTc Time QTc Drug Placebo True Treatment response Time QTc Time QTc Drug Placebo Baseline QTc Drug Time QTc Placebo Time QTc
34 Step 3: Estimate Response - - Estimated Treatment Effect = = e.g. Sampled Baseline Time QTc Time QTc Examples of approaches to estimating treatment effect Mean(sampled response): treatment - baseline Max(sampled response): treatment - baseline # Subjects with outlying values: treatment - baseline e.g. Sampled Response Time QTc Time QTc Drug Placebo
35 Step 4: Repeat Many Times Randomly pick baseline data ↓ Simulate response to treatment ↓ Estimate drug effect QTcF Study 15 msec Study 28 msec … Study msec
36 Step 5: Evaluate Performance e.g. Power Fraction of simulations in which Ho rejected Ho: mean change from baseline for drug = placebo
37 Preliminary Results
38 QT baseline source data Resample from 72 hr QTc data in 45 subjects Individual Trial Design Randomized, parallel, with treatment and placebo arms 24 hour placebo run in; 24 hours on treatment Hourly QT sampling from 1-24 hours N: varied Treatment characteristics p.o. administration Dose = 100 mg PK: One compartment; ka: 1/hr, CL: 25 L/hr, V: 400 L, Tmax: 2-5 hours PK/PD: Linear relationship, no effect delay Drug effect on QTc is additive on top of the baseline variation Intersubject variability of PK and PK/PD are 25% Analysis 24-hr max QTc during treatment – 24-hr max QTc for baseline 24-hr mean QTc during treatment – 24-hr mean QTc for baseline 24-hr max QTc during treatment – 24-hr average QTc during baseline 24-hr max QTc during treatment (no baseline subtraction) Study Results: 200 clinical trials simulated
39 Time (hr) log (Conc ng / mL)) Tmax: 2-5 hours QTc from Baseline (msec) Time (hr) Rmax ~ 16 msec Treatment Characteristics
40 Power to Detect a Difference Between Drug and Placebo with 95% Confidence
41 Several QT samples at baseline One QT sample at baseline Effect of Baseline Measurement on Power
42 Questions for the committee 1. What additional study design points are recommended for consideration in the analysis of PK-QT data? 2. Comment on the case studies presented and the pros and cons of using clinical trial simulation approaches to evaluate PK-QT study design. Are there other methods of analyzing PK-QT data that the FDA should consider? 3. What critical design elements influence the outcome of a PK-QT study that has as its goal to identify a meaningful change in QT?