Transcript Classification of clinical trials

Classification of clinical trials
Chapter 7 Reading instructions
•
•
•
•
•
•
•
•
•
7.1 Introduction
7.2 Multicenter Trials: Read + extra material
7.3 Superiority Trials: Read
7.4 Equivalence/Non-inferiority Trials: Read
7.5 Dose Response Trials: Read
7.6 Combination Trials: Read
7.7 Bridging Trials: Skip
7.8 Vaccine Trails: Skip
7.9 Discussion
Multicenter trials
All large studies are conducted at multiple centers.
Center=clinic=study site
Issues:
• Treatment by center interaction*
• Estimation of treatment effect.
*) Covered in the lecture on basic statistical concepts
Multicenter trials
Estimation of treatment effect
Model: effect=center + treatment + center*treatment + error
Sum of squares:
Type I
Varation due to each specific
factor beyond those already in
the model according to order
of specification.
Type II
Varation due to each specific
factor beyond what can be
explained by the others,
excluding interactions with
this specific factor
Type III
Varation due to each specific
factor beyond what can be
explained by all other
specified factors including
interactions
Multicenter trials
Assume k centers with
True treatment effect:
i
Estimated treatment effect:
ˆi
Variance of estimated treatment effect:
k
Overall treatment effect estimated by
 w ˆ
i 1
Type II estimator: wi 
1 /  i2
i 1
Treatment effects averaged over
center weighted according to
precision (think ni).
k
where
w
i 1
Type III estimator:
k
1 / 
i i
 i2
2
i
Treatment effects averaged
over center with equal
weight for all centers
i
wi  1
1
k
Multicenter trials
The effect of center imbalance on type III estimates and test
80
1
0.9
70
0.8
60
50
0.6
40
Standard error of
estimated treatment effect
30
Power of treatment
comparison
0.5
Power
Standard error
0.7
0.4
0.3
20
0.2
10
0.1
0
0
0
10
20
30
40
50
60
Percentage of patients on the smallest of two centers
200 patients split on 2 centers, response N(50,25) and N(80,25), no true center effect or
interaction between center and treatment.
Model including treatment, center and treatment*center, simulated 1000 times.
Superiority, equivalence and
non-inferiority
Experimental treatment with true mean effect: T
Control treatment with true mean effect:
Superiority:
Equivalence:
Non-inferiority:
C
The experimental treatment is better
than the control treatment.
The experimental treatment and the
control treatment are similar.
The experimental treatment is not that
much worse than the control treatment.
H0 : T  C
H1 : T  C
H0 : T  C  d
H1 : T  C  d
H 0 : T  C  d
H1 : T  C  d
Superiority
Superiority:
The experimental treatment is better
than the control treatment.
The experimental treatment is not
better than the control treatment
H0 : T  C
H0 : T  C
H1 : T  C
The experimental treatment is
better than the control treatment
0
T  C
H1 : T  C
REJECTED
0
0 ˆT  ˆ C
ˆT  ˆ C
T  C
95% conf. Int.
T  C
Equivalence
The experimental treatment and the
control treatment differ at least d
The experimental treatment and the
control treatment differs less than d
H : T  C  d

0
-d
0
H : T  C  d

0
d
T  C
The combined null hypothesis H0 can be tested at level  by
testing each of the two disjunct components also at level .
REJECTED
H : T  C  d

0
REJECTED
-d
0
T  C
d
H : T  C  d
90% conf. Int.

0
T  C
What is similarity?
Example:
We have developed a new
formulation (tablet) for our
old best selling drug. How
do we prove that the new
formulation has the same
effect as the old one
without new big studies?
Due credit to Chris Miller AstraZeneca biostatistics USA
Bioequivalence
FDA definition:
Pharmaceutical equivalents whose rate and extent of
absorption are not statistically different when administered
to patients or subjects at the same molar dose under similar
experimental conditions
Operationalized as:
Compared exposure in terms of AUC and Cmax of
the plasma concentration vs time curve and
conclude bioequivalence if the confidence for the
ration between the formulations lies between 0.8
and 1.25 for both AUC and Cmax.
Concentration (nmol/L)
Crash pharmacokinetics
Time of dose=0
AUC=Area Under Curve
Distribution: The drug is distributed to the
various sub compartments of the
body.
Elimination: The drug leaves the circulation by
metabolism or excretion at an
exponential rate.
Time (h)
Absorbtion: The drug is absorbed from the
cite of administration leading to
increased concentration in the
blood (plasma).
Cmax=maximal concetration
Tmax=time to Cmax
Example
A phase I, open, randomized, four-way cross-over, single-centre study to
estimate the pharmacokinetics and tolerability of single oral doses of 100 mg
AR-H044277XX given as two different mesylate salt tablets, a base form tablet
and an oral solution in healthy male subjects.
The primary objectives of this study are to estimate and compare the
pharmacokinetics of two different mesylate salt tablets, a base form tablet and an
oral solution given as single oral doses of 100 mg AR-H044277XX in healthy male
subjects by assessment of AUC and Cmax.
Design:
A
B
C
D
B
D
A
C
C
A
D
B
D
C
B
A
Model: Mixed effect ANOVA
Yijk     i  i ( k )   j   t   ijk
x
Example cont.
8
Mean plasma concentration (µmol/L)
7
6
micr base form
5
mesylate salt
4
micr mesylate salt
3
oral solution
2
1
0
0
4
8
12
16
20
24
28
32
36
Time af ter dos e (h)
Table 1
Parameter
AUC
Cmax
Geometric mean of the ratio between XX/sol, AW/sol and AW micr/sol
of AUC, AUCt and Cmax (n=13)
Ratio
XX/sol
AW/sol
AW micr/sol
XX/sol
AW/sol
AW micr/sol
Estimate
0.95
1.08
1.08
0.97
1.04
1.04
90% CI
lower
0.85
0.90
0.96
0.84
0.90
0.78
upper
1.07
1.32
1.22
1.13
1.20
1.39
XX and sol are bioequivalent by not AW and sol or AWmicr and sol.
Non inferiority
The experimental treatment and the
control treatment differ at least d
The experimental treatment and the
control treatment differes less than d
H0 : T  C  d
-d
T  C
0
T  C
REJECTED
H0 : T  C  d
-d
0 ˆT  ˆ C
95% conf. Int.
Why Non inferiority?
Intrinsic non-inferiority:
To claim that the effect of the test drug is at least not
to a relevant degree worse than the comparator.
Often combined with superiority on other variable
e.g. non inferior effect and superior safety.
Indirect superiority:
To claim that the effect of the test drug is
better than placebo when placebo not
considered ethical.
Dose Response trials objectives
Objectives: • Confirm efficacy
• Investigate shape of dose reponse curve
(ICH E9)
• Estimate a appropriate starting dose
• Indentify optimal individual dose adjustment strategy
• Determination of maximal dose
•Safety!
Beneficial effect
Effect
emax
e50
Toxic effect
e0
ed50
Dose
Therapeutic window
Dose
Concentration(s) Effects
c
E
Time
One dose results in concentration
vs time profiles for the given
compound as well as one or
several metabolites
Time
Depending on the mechanisms
of action we get effect vs time
profiles for both wanted and
unwanted effects
Inter individual variation in dose response
Effect
Population average
Dose
Dose response trials design
Design options
Parallell groups
Forced titration
•Large
•Easy to impement
•No confounding
•Easy to analyse
•Small
•Confounding
Cross over
•Small
•Carry over effects
Dose response trials design
Dose 1
Design:
Dose 2
R
N=n1
N=n2
Parallell groups with k dose
groups and a control.
Placebo?
Dose k
Control
N=nk
N=nk+1
How are the doses selected?
Which sample size(s) should
be used?
Dose Response trials models
Separate means with equal variance
yij  i   ij
;
 ij iid N 0, 2 
Regression model
Effect
emax
Emax  d 
E  E0  
ed50  d 
e50
e0
ed50
Dose
Confiriming efficacy
No placebo but
significant linear dose
response means efficacy
confirmation
Dose 1 Dose 2
Dose 3
Significant difference
from placebo confirms
efficacy
Placebo Dose 1 Dose 2 Dose 3
Noninferiority to active
control confirms efficacy
Active Dose 1 Dose 2 Dose 3
Dose response trials design
options
Fixed design: doses and sample sizes are descided upfront
and subsequently not changed during the trial.
Adaptive design: Initial doses and sample sizes are
descided upfront may subsequently be changed during the
trial depending on the outcome.
Figure 4
4
3
ongoi ng /
outcome
patient data
dose to vial
translation
2
New
patient
• Bayesian
• Frequentist (D-optimality)
6
estimation
of dose-response
curve
randomisation
to pl acebo or
“optimal ”dose
1
Basic adaptive variants:
5
longitudi nal
model predi cts
fi nal outcome
deci si on
rul e
fi nd the
optim al dose
for learni ng
about ED95
Continue
8
Stop
10
Go Piv otal
11
7
9
Dose response trial analysis
options
Pairwise comparisons: The doses are compared using
significance tests often adjusted for multiple comparisons and the
aim is to show effect vs the comparator and to separate the
doses.
•
•
•
•
Limited assumptions
Easy to compare doses
Need relatively many observations
No estimate of a dose reponse curve
Model based: The effect is assumed to follow a parameteric
model with parameters estimated from the data.
•
•
•
•
More assumptions
Tricky to compare doses
Need relatively few observations
Estimates a dose reponse curve
Example of the design of a
dose response trial
We could also find litterature
data relating the effect on the
biomarker to clinical effect.
the percentage of time with intragastric pH>4
based on data from C2, C6, C18 and L4
100
Percentage of time with pH>4, 0-24h (%)
In this example we have plenty
of data on the relation
between the dose and the
effect on a biomarker.
Predicting
90
80
70
60
50
40
30
Modelled % time with pH>4
20
Observed % time with pH>4
10
0
0
20
40
60
80
100
120
140
160
Dose AZDXXXX (mg)
Predicting the 4 week healing rate using a log linear
model
Percentage healed after 4
weeks
The design of a dose response
trial is based on data from
previous studies with the same
or similar drugs.
90
80
70
60
50
40
30
20
10
0
P40
L30
E40
E20
O20
8
9
10
y = 20.26Ln(x) + 22.143
11
12
13
14
15
16
Tine with pH>4 (h)
17
18
19
20
21
22
Example of the design of a
dose response trial
Samplesize:
•Highest dose to have power 80% against competitor
•Same sample size in all groups
Combination treatments
Evaluation of a fixed combination of drug A and drug B
A placebo A active
B placebo B placebo
A placebo A active
B active
B active
P
A
B
AB
The U.S. FDA’s policy (21 CFR 300.50) regarding the use of a fixed-dose combination
The agency requires:
Each component must make a contribution to the claimed effect of the
combination.
Implication: At specific component doses, the combination must be superior to its
components at the same respective doses