Exposure-AE-Dropout Analysis in Patients treated with pregabalin.

1
Exposure-AE-Dropout Analysis in Patients treated
with pregabalin.
Pfizer Global Research and Development

• Raymond Miller

2
Issue

• A new ?2? ligand (PD0332334) that has anxiolytic
  properties was in development.
• Little was known about AEs for this compound,
  however, extensive knowledge from other ?2?
  ligands (pregabalin) available.
• It is generally believed that dose titration may
  reduce AEs.

3
Objective

• To characterize the relationship between
  PD0332334 dose, patient characteristics, time,
  severity and frequency of dizziness and
  somnolence in patients with GAD.

4
Questions

• Would AE frequency be different if the drug was
  titrated to the target dose?
• How long do we need to titrate to minimize AEs?
• How many dose steps do we need to minimize AEs?

5
Current Information

• Multiple phase 3 trials with pregabalin titrated
  over 3 to 7 days to attain steady state dose in
  the treatment of GAD.
• One phase 4 study with three treatment groups
  placebo, pregabalin 600 mg/day fixed, 150-600
  mg/day titrated..

6
Phase 3 trialsGAD patients treated with
Pregabalin

• 1630 patients information (47218 observations)
  was pooled from 6 clinical studies.
• All studies consisted of treatment arms with a
  dose titration phase varying from 3 to 7 days
  followed by a three or five week maintenance.
• Dizziness was spontaneously recorded using a
  daily diary as none0, mild1, moderate2, and
  severe3.
• Dropout was recorded as such up to 3 days before
  scheduled conclusion of the study.

7
Objectives

• To describe the exposure-longitudinal AE severity
  relationship following multiple doses of
  pregabalin.
• To describe the relationship between AE and
  patient dropout
• To explore the relationship between dose
  titration of pregabalin and dropout

8
Frequency of dizziness by day and dose
9
Exposure-Dizziness-Dropout in GAD patients
treated with Pregabalin

• Models were developed for exposure-AE as well as
  AE-dropout.
• For AE separate models were developed for the
  incidence of adverse event and for the
  conditional severity of adverse event given that
  an adverse event has occurred.
• The unconditional severity probability
  distribution was obtained by summing the joint
  probabilities.
• Dropout was modeled using a discrete time
  survival model.

10
Assumption that ?j Niid(0, ?2) is violated.
11
Incidence Model

• The probability of incidence of dizziness was
  modeled using a nonlinear logistic regression
  model given by the expression
• The incidence model does not contain an
  inter-individual random effect because AEi is
  observed only once for each patient
• Sigmoid Emax model best describes the drug effect
  although ? is not well estimated

12
Observed vs. PredictedIncidence Model
13
Conditional Severity Model

• The probability of each severity (none, mild,
  moderate, severe) was modeled with a proportional
  odds model. The conditional severity model given
  by the expression
• Drug exposure was based on the intended daily
  dose (titrated) of pregabalin.
• Emax model with time-course placebo effect and a
  component with an exponential attenuation best
  describe the AE severity.

14
Dataset and NONMEM control stream
PRED B1THETA(1) B2B1THETA(2)
B3B2THETA(3) logits for Ygt1, Ygt2,
Y.3 RESP0 A1 B1 RESP ETA(1) A2 B2
RESP ETA(1) A3 B3 RESP ETA(1)
C1EXP(A1) C2EXP(A2) C3EXP(A3) probabilities
for Ygt1, Ygt2, Ygt3 P1C1/(1C1)
P2C2/(1C2) P3C3/(1C3) Probabilities for
Y0 Y1, Y2, Y3 PA1-P1 PBP1-P2 PCP2-P3
PDP3
15
Observed vs. PredictedConditional Severity Model
16
Markov Model

• Markov elements have been incorporated to account
  for the correlation between neighboring
  observations within a subject
• The logistic function (proportional odds model)
  and the same structures obtained with the
  conditional severity model was used.

17
Dataset and NONMEM control stream
PRED B1THETA(1) B2B1THETA(2)
B3B2THETA(3) IF(PRE1.EQ.1) THEN B1THETA(4)
B2B1THETA(5) B3B2THETA(6) ENDIF IF(PRE1.EQ.2
) THEN B1THETA(7) B2B1THETA(8)
B3B2THETA(9) ENDIF IF(PRE1.EQ.3) THEN
B1THETA(10) B2B1THETA(11) B3B2THETA(12) END
IF RESP0 A1 B1 RESP ETA(1) A2 B2
RESP ETA(1) A3 B3 RESP ETA(1) .. ..
18
Observed vs. PredictedConditional Severity Model
with Markov
19
Simulation Step(example Time-course of
incidence)
ID1
Probability of Incidence
Mean of trial
ID2
N times simulations
ID3
Summary of Mean
..
ID1630
Original Dataset
20
Posterior Predictive Check Distributions of the
Number of the Different Transitions
with Markov
without Markov
The vertical line in each plot represents the
observed number of transition in the original
dataset
21
Simulation (mild)Severity Model with Markov
Simulated Probabilities Are Presented By Means
(lines) with 95 CI (dash lines) and 80 CI
(shades) from 100 Simulations.
22
Simulation ( moderate)Severity Model with Markov
Simulated Probabilities Are Presented By Means
(lines) with 95 CI (dash lines) and 80 CI
(shades) from 100 Simulations.
23
Simulation (severe)Severity Model with Markov
Simulated Probabilities Are Presented By Means
(lines) with 95 CI (dash lines) and 80 CI
(shades) from 100 Simulations.
24
Conclusion

• The probability of experiencing dizziness during
  any day increases with pregabalin daily dose.
• The predicted mean incidence of dizziness was
  around 35 at daily dose of 200 mg/day or
  greater, which was at least 2 fold higher
  compared to those of at daily doses lt150 mg/day.
• The most frequently reported severity was mild to
  moderate. The risk of experience dizziness with
  any severity increases within 1 week, but decline
  to over the next 3 to 4 weeks. The risk of mild
  or moderate dizziness increases up to 25 within
  1 week, and declines to around 7 over 3 to 4
  weeks.
• The proportional odds model including a Markov
  element could describe the time-course of
  probability of dizziness well.

25
Dropout Model
26
Dropout Across Doses
27
Dropout Model

• Dropout was modeled using a discrete time
  survival model (Gompertz).
• Dizziness severity was included in the model as a
  covariate.

g(w,Yt-1) represents the hazard function
28
Simulations of dropout probabilities based on
simulated severity of dizziness stratified by
representative unique dose titration profiles
over time. Observed (red line)
29
GOF 5th 95th prediction interval constructed
from 200 simulations using the original dataset
structure as well as median model predicted
dropout (grey line) and Kaplan-Meier estimates of
in study-survival (black line).
30
External ValidationPregabalin BID Add-On
Titration Trial A Randomized, Double-Blind,Place
bo-Controlled, Parallel-Group, Multicenter Study
in Patients With Partial Seizures (1008-157)
31
TIME TO WITHDRAWAL
32
External ValidationObserved (Kaplan Meier)
dropout from an independent 12 week GAD trial
(red line) with either placebo or 600 mg daily
pregabalin treatment and its corresponding
5th-95th nonparametric confidence intervals at
weekly increments. Gray polygon outlines a
prediction interval of 5th and 95th quantiles of
1000 trial simulation using the described GAD
dropout model
33
Titration Scenarios300 mg daily ITT
Scenario 1 (1week) 50x2, 100, 150, 200, 250,
300 Scenario 2 (2week) 50x3, 100x3, 150x2,
200x2, 250x2, 300... Scenario 3 (3week) 50x4,
100x4, 150x4, 200x4, 250x3, 300... Scenario 4
(4week) 50x6, 100x5, 150x5, 200x5, 250x5,
300... Scenario 5 (6week) 50x8, 100x8, 150x8,
200x8, 250x8, 300...
34
gtmild
35
gtmoderate
36
gtsevere
Note y-axis scale is adjusted to enlarge the AE
profile
37
Simulated GAD survival probabilities from the
combined Dizziness-dropout model. Two dosing
schemes (blue) within a weeklong titration
regimen differ only over 3 initial days of
dosing.
38
Next Steps

• Clinical Trial Simulations using different
  titration scenarios.
• Titration over different times
• Variations in the first week.
• Scaling to drugs in same class

39
Acknowledgements

• Kaori Ito
• Bojan Lalovic
• Matt Hutmacher

40
Backup
41
gtmild
42
gtmoderate
43
gtsevere
Note y-axis scale is adjusted to enlarge the AE
profile
44
(No Transcript)
45
Simulation of dropout
46
Graphical Data Exploration- Nonparametric/Kaplan
Meier Analysis Poolability of Placebo Cohorts GAD
number of events
number at risk
At ti there are di events (dropouts) and ni
individuals (at risk).
In Splus (survfit) only accommodates categorical
time-invariant covariates (strata)!
47
Comparison of Dropout Across Titration Schemes GAD
RD_at_ Day 6 100 mg/day
RD_at_ Day 7 150 mg/day
600 mg
400 mg
Longer titration (time-to-attainment of
randomized dose) -gtlower dropout
Study 87- an outlier?