Download SAS application: PEST 4 for sequential clinical trials

SAS application:PEST 4 for
sequential clinical trials
Tatsuru Matsushita
Sequential Clinical Trials
•
•
•
•
•
Prove efficacy of a drug
Observe its action in human subjects
Need to minimize the number of patients
Reduce the number by testing bit by bit
Hence, ‘sequential’ clinical trials
Triangular design
• Stop trial as soon as efficacy is proved,
or non-efficacy is established
Simulation condition
• Specifying the way to generate patients data
Single simulation
• Simulating the trial using random data
• Assuming the drug has planned efficacy
CONTINUE the study - no boundary has yet been crossed
Estimated sample size
• Necessary number of patients for various
efficacy of the drug
Variable(s) plotted against theta :
Med(n*) P90(n*)
180
170
160
150
140
130
120
110
100
90
80
70
60
50
40
-2
-1
Variables plotted:
0
theta
Med(n*)
1
2
P90(n*)
Various aspects of SAS
• Use 150 SAS/AF frame/scl entries
• Convert C code into SAS procedure using
SAS/TOOLKIT
• Use PROC LOGISTIC, PROC PHREG, and
PROC IML
Summary
• Sequential method in clinical trials has
potential to reduce the number of patients.
• Simulation facility enables to test and refine
trial designs without recruiting patients
• By specifying recruitment schedule the
probability of reaching a firm conclusion
can be estimated in survival type trials
• Various SAS modules, procedures are used
Table 1. Results from multiple simulation
Simulation of <C:\HOME\MY DOCUMENTS\SEUGI2002\BINARY>
Number of simulations = 10000
Value of theta in simulation =
1.09861
Coverage probability of confidence intervals = 0.95
Continuous analysis
Boundary crossed :
upper
lower
Final p-value :
count (prop.)
8960 (89.6 %)
1040 (10.4 %)
count (prop.)
p <= alpha & E superior
p > alpha
p <= alpha & E inferior
8960 (89.6 %)
1037 (10.4 %)
3 ( 0.0 %)
Terminal statistics
_____________________________________________________________
Z
V
# INTERIMS
_____________________________________________________________
median
6.13333
5.28889
9.00000
P90.0
7.50000
8.63923
15.00000
mean
5.95667
5.54668
9.55890
std dev.
1.64783
2.17149
3.74456
_____________________________________________________________
Sample sizes at termination
_____________________________________________________________
NE
NC
N
_____________________________________________________________
median
45.00000
45.00000
90.00000
P90.0
75.00000
75.00000
150.00000
mean
47.84350
47.85350
95.69700
std dev.
18.76909
18.75931
37.46992
_____________________________________________________________
Success rates (weighted averages)
_____________________________________________________________
E
C
Overall
_____________________________________________________________
average
0.49903
0.24991
0.37446
std error
0.00072
0.00063
0.00049
_____________________________________________________________
Table 2. Design Properties
theta/theta_r
-1.50
-1.00
-0.50
0.00
0.50
1.00
1.50
p(con)
p(exp)
0.25
0.25
0.25
0.25
0.25
0.25
0.25
0.06
0.10
0.16
0.25
0.37
0.50
0.63
E(n*)
53.73
61.26
72.89
95.17
118.42
91.91
57.25
Med(n*)
49.66
56.29
66.21
85.79
114.22
84.54
51.11
P90(n*)
71.65
84.87
106.46
149.45
176.85
143.73
85.28
Table 3. Simulation runs assuming non efficacy of the drug
Simulation of <C:\HOME\MY DOCUMENTS\SEUGI2002\BINARY>
Number of simulations = 10000
Value of theta in simulation =
0.00
Coverage probability of confidence intervals = 0.95
Continuous analysis
Boundary crossed :
upper
lower
Final p-value :
count (prop.)
238 ( 2.4 %)
9762 (97.6 %)
count (prop.)
p <= alpha & E superior
p > alpha
p <= alpha & E inferior
239 ( 2.4 %)
9491 (94.9 %)
270 ( 2.7 %)
Terminal statistics
_____________________________________________________________
Z
V
# INTERIMS
_____________________________________________________________
median
-0.27
1.74
4.00
P90.0
1.84
2.96
7.00
mean
-0.00
1.92
4.13
std dev.
1.40
0.77
1.76
_____________________________________________________________
Sample sizes at termination
_____________________________________________________________
NE
NC
N
_____________________________________________________________
median
20.00
20.00
40.00
P90.0
34.00
34.50
70.00
mean
21.06
21.08
42.14
std dev.
8.99
9.00
17.88
_____________________________________________________________
Success rates (weighted averages)
_____________________________________________________________
E
C
Overall
_____________________________________________________________
average
0.25
0.25
0.25
std error
0.00
0.00
0.00
_____________________________________________________________
Table 4. Simulation which runs out of patients
Simulation of <C:\HOME\MY DOCUMENTS\SEUGI2002\SURVIVAL>
Number of simulations = 1000
Value of theta in simulation =
0.92
Coverage probability of confidence intervals = 0.95
Continuous analysis
Boundary crossed :
upper
lower
Final p-value :
count (prop.)
701 (70.1 %)
70 ( 7.0 %)
count (prop.)
p <= alpha & E superior
p > alpha
p <= alpha & E inferior
799 (79.9 %)
201 (20.1 %)
0 ( 0.0 %)
Terminal statistics
_____________________________________________________________
Z
V
# INTERIMS
_____________________________________________________________
median
7.05
7.72
6.00
P90.0
8.57
10.20
9.00
mean
6.68
7.55
6.35
std dev.
2.14
2.16
1.90
_____________________________________________________________
Sample sizes at termination
_____________________________________________________________
NE
NC
N
_____________________________________________________________
median
55.00
55.00
110.00
P90.0
56.00
56.00
110.00
mean
54.98
54.96
109.94
std dev.
1.22
1.12
1.09
_____________________________________________________________
Number of events at termination
_____________________________________________________________
# EVENTS
_____________________________________________________________
median
31.00
P90.0
41.00
mean
30.49
std dev.
8.62
_____________________________________________________________