Transcript Sample Size Estimation Incorporating Disease Progression

School of Public Health
Sample Size Estimation
Incorporating Disease
Progression
Barbara C. Tilley, Ph.D.
Keith Burau, Ph.D.
University of Texas, School of Public Health
Jordan Elm, Ph.D.
Medical University of South Carolina
Acknowledgment
This presentation is based on:
Guimaraes P, Kieburtz K, Goetz CG, Elm JJ,
Palesch Y, Huang P, Ravina B, Tanner CM,
Tilley BC. Non-linearity of Parkinson’s
disease progression: implications for sample
size calculations in clinical trials.
Clinical Trials 2005; 2:509-518
Parkinson’s Disease (PD) Issues
• Many short term studies enroll
untreated (de novo) patients.
• De novo patients are followed to
dopaminergic therapy and LOCF is
used, or to study end point, whichever
occurs first.
• Once a patient starts dopaminergic
therapy there is a decrease in the
manifestions of PD.
UPDRS scores
• The gold standard PD rating scale
• Measures combine questions on Motor,
ADL, mental function
• Purpose of Guimaraes et al’s approach
was to characterize progression of
UPDRS scores after initiation of
symptomatic therapy
UPDRS Scores
Evolution of UPDRS Scores
Start Sx Tx
3 - 6 months
Time
5 years
UPDRS Scores
Long Term Follow-up of UPDRS scores
with Different Study Populations
Baseline
End of Study
STUDY TIME
Sample Size Estimation
• Based on data from other PD clinical trials
(CALM-PD and GSK) we developed a non-linear
model of the symptomatic effect of
antiparkinsonian medications on disease
progression as measured by UPDRS scores.
• We used the model to:
– Predict change in Total UPDRS;
– Predict the variance (s.d.) associated with
change in Total UPDRS;
– To infer sample size
Non-linear Model
Yi(t) = b0i + b1it - b2i (1-eb3t) + it
• The model assumes that scores reflect
the sum of two components:
• A linear trend reflecting disease
progression
• A symptomatic effect that “wears-off”
over time
Si(t)=b2i (1-eb3t)
BASELINE = Initiation of Levodopa
Estimated Changes in Total UPDRS
(Based on CALM-PD data)
Year
1
2
3
4
5
Change
from
Baseline
-12.4
-9.6
-6.9
-4.1
-1.4
30% Change
from
Baseline*
-3.7
-2.9
-2.1
-1.2
-0.4
Standard
Deviation of
Total Change
10.22
11.52
13.83
16.75
20.00
* 30% or 3 points change in UPDRS over over 1 yr = success
BASELINE = Initiation of Levodopa
Sample-Size Requirements
(Based on CALM-PD data)*
Year
1
2
3
4
5
Alternative 1
(30% relative change)
160
338
940
3899
47657
Alternative 2
(3 point change)
245
311
448
657
935
* Assumes actual value used, not LOCF
BASELINE = 6 months after Levodopa
Estimated Changes in Total UPDRS
(Based on CALM-PD data)
Year
1
2
3
4
5
Estimates for Total UPDRS
Change 30% Change
Standard
from
from
Deviation of
Baseline
Baseline
Total Change
2.6
0.8
7.03
5.4
1.6
9.74
8.1
2.4
13.07
10.9
3.3
16.65
13.7
4.1
20.34
BASELINE = 6 months after Levodopa
Sample-Size Requirements
(Based on CALM-PD data)*
Year
1
2
3
4
5
Alternative 1
(30% relative change)
1709
761
609
546
516
Alternative 2
(3 point change)
117
223
400
649
967
* Assumes actual value used, not LOCF.
Another Issue in PD
• Reduction in time from diagnosis to
initiation of antiparkinson/dopaminergic
therapy (due to clinical practice) is
adding to difficulty of studying untreated
(de novo) patients
Kaplan-Meier curve of time from diagnosis
to need for symptomatic therapy.
100
CoQ10
GPI-1485
Current Placebo
Creatine
Minocycline
Prior Placebo
DATATOP Placebo
90
80
70
60
50
40
30
20
10
0
0
1
2
3
4
Years from Diagnosis
5
6
14
Sample Size for NET-PD LS-1
• Objective: Determine if there is > 1 year *
difference in slowing of clinical decline over
5-year period comparing creatine treated
participants to placebo treated participants.
• Inclusion: All participants currently on
dopaminergics drugs at least 90 days but <
2 years
*1.5 years for ambulatory capacity
Sample Size
• Calculations based on data collected in other
studies for patients on dopaminergic
treatment at least 90 days.
• Clinical decline: change from baseline* over 5
years using a combination of 5 cognitive,
physical, and quality of life measures
combined into a single outcome and
compared with a global statistical test
(O’Brien Ranked Test).
* Rankin – final value
Improvement in Long Term
Progression
Statistical
Test of
•Global
Ambulatory
capacity
• Ambulatory capacity (5 UPDRS Questions)*
• Modified Rankin (functional disability scale)
• Modified Rankin (functional disability scale)
• PDQ-39 (Quality of Life)
• PDQ-39 (Quality of Life)
•• Schwab
and
England
(ADL
scale)
Schwab and England (ADL scale)
•• Symbol
DigitModalities
Modalities(measure
(measureofof
cognitive
Symbol Digit
cognitive
impairment)
impairment)
*Target of dopaminergic medications
LS-1 Sample Size (cont)
• N= 860 per group
• Power > 85% to detect a 1-1.5 year slowing
of clinical decline in the treatment arm
compared to control
• Chose largest sample size for a single
outcome
• alpha = 0.05, 2-sided.
• Adjusted for 20% drop-ins and drop-outs with
inflation factor
• Overall GST gives > 90% power
Considerations in Long Term
Study
• Avoids using short term LOCF value for long
term outcome
• Avoids non-linearity issue identified by
Guimaraes.
• Objective is improvement in long-term clinical
decline rather than distinguishing
symptomatic from disease modification
effects
• Expected to take at least 5 years to see
changes in outcomes other than UPDRS.
Delayed Start Design for Chronic
Disease
Phase 1
Phase 2
Disease
Mod.
Effect
A
P
A
A
0
T
Baseline 1
Time Point
T
2
T
3
T
4
Delayed Start
• Goal to separate symptomatic from disease
modifying effects
• Provides shorter follow-up on placebo
• More assumptions (e.g. duration of
symptomatic effect)
• Can be affected by non-linearity if de novo
patients included.
• Sample size implications if placebo goes on
treatment
NET-PD FS-Zone Phase II
Futility Study of Pioglitazone
• Inclusion criteria – on stable dose of either
1 mg/day of rasagiline or 10mg/day of
selegiline for 8 months before
randomization.
• Expected to remain on the fixed dose for
the duration of the trial.
• Sample size - 72 pt per group to detect a
3 point change in UPDRS, two-sided alpha
= 0.15, power = 0.85.
FS-ZONE
• Baseline examination roughly coincides
with the period in which the patient is
experiencing the maximum symptomatic
benefit of the concurrent antiparkinson
medication.
• Thus the nonlinearity characterized by
Guimaraes et al 2005 is avoided.
Considerations for Phase II Studies
of Shorter Duration
Starting de novo patients on antiparkinson
therapy before trial will:
– Reduce variability
– Reduce potential confounding (of other
antiparkinsonian meds)
– Reduce missing data (patients dropping
out to go on other treatments)
– Increase likelihood of recruitment.
Alzheimer’s Disease
Report of the task force on designing clinical trials
in early (predementia) AD (e-Pub ahead of print).
Neurology 2011.
Painful Joint Scores
MIRA Joint Pain Score
Treatment Groups Contrasting Past or de novo DMARD use
Weeks of Follow Up
Treatment
* Nonlinear Model Fit to 48 weeks of Observations
Conclusions
• Parkinson’s Disease (PD)
– Effects of l-dopa and other dopminergic
therapies on decline, and course of PD,
and Guimaraes model suggest considering
the effect of dopaminergic therapy on
decline is essential in the design, at least
for UPDRS (standard outcome measure).
Conclusions Continued
• Early Alzheimer’s Disease
– Shares some characteristics of PD in terms
of decline
– Figure presented does not address
introduction of other therapies
– Therapies such as Aracept and others that
may slow or mask decline may create
similar design issues.
– It would be useful to model Alzheimer’s
data using Guimaraes approach
Conclusions Continued
• Rheumatoid Arthritis (RA)
– Course of disease more erratic due to
flares, remissions
– Immediate and long term effects of
DMARDs vary
– Trials usually require stable dose of
prednisone, methotrexate as entry criteria.
– More analyses of RA data sets with longer
follow-up and stronger DMARDS would be
of interest; effects on sample size less
likely
EXTRA SLIDES
LS-1Sample Size (1-Yr Diff. in Progression)
Measure
5 year chg.
PDQ-39
Summary score
Minimum
Expected detectable Common Effect Source
Placebo Difference SD
Size
mean
in means
15
3
9
0.33 Olanow
et al. 2004
Modified Rankin
(final value)
2.2
0.2
1
0.20
FS1, FS2,
and
K. Shannon
(unpublished)
Schwab &
England ADL
10
2
11
0.18
PEP/PEPX
Kieburtz et
al. 1994
Symbol-Digit
Modalities
Ambulatory
Capacity
5.5
1.5
8.0
0.19
1.25
0.38
2.11
0.18 CALM-PD