Transcript 20080724140014302

Pharmacogenetics - difficult or
just impossible?
Stephen Senn
(c) Stephen Senn 2007
1
Based on chapter 25 (with
some additional material from
chapter 24).
(c) Stephen Senn 2007
2
“Statistics and the medicine of the
future”
Mass-market drugs have successfully treated millions, but
they have a corollary: one size has to fit all. Every patient gets
the same drug – yet every patient is different and responds
differently to drugs, treatments and doses…Each drug each
dose, each treatment will be tuned not to the average patient
but to the individual. It is the difference between an off-thepeg suit and one made to measure.
Chris Harbron, Significance, June 2006, p67 (My italics)
3
Genes, Means and Screens
It will soon be possible for patients in clinical trials to undergo genetic
tests to identify those individuals who will respond favourably to the drug
candidate, based on their genotype, and therefore the underlying
mechanism of their disease. This will translate into smaller, more
effective clinical trials with corresponding cost savings and ultimately
better treatment in general practice. In addition, clinical trials will be
capable of screening for genes involved in the absorption, metabolism
and clearance of drugs and the genes which are likely to predispose a
patient to drug-induced side-effects. In this way, individual patients will
be targeted with specific treatment and personalised dosing regimens to
maximise efficacy and minimise pharmacokinetic problems and other
side-effects.
Sir Richard Sykes, FRS
(c) Stephen Senn 2007
4
Claims for
Pharmacogenomics
• Clinical trials
– Cleaner signal
– Non-responders eliminated
• Treatment strategies
– “Theranostics”
• Markets
– Lower volume
– Higher price per patient day
(c) Stephen Senn 2007
5
Pharmacogenetics: A cutting-edge
science that will start delivering miracle
cures the year after next.
(c) Stephen Senn 2007
6
Implicit Assumptions
• Most variability seen in clinical trials is genetic
– Furthermore it is not revealed in obvious phenotypes
• Example: height and forced expiratory volume (FEV1) in one second
• Height predicts FEV1 and height is partly genetically determined but
you don’t need pharmacogenetics to measure height
• We are going to be able to find it
– Small number of genes responsible
– Low (or no) interactive effects (genes act singly)
– We will know where to look
• In fact we simply don’t know if most variation in clinical
trials is due to individual response let alone genetic
variability
(c) Stephen Senn 2007
7
Moerman and Placebos
• Paper of 1984
• Investigated 31 placebo-controlled trials of
cimetidine in ulcer
• Found considerable variation in response
• Considered placebo response rate was an
important factor
• Has been cited by others as proof of
variation in treatment effect from trial to
trial
(c) Stephen Senn 2007
8
31 Placebo-Controlled Trials of Cimetidine
4
log-odds ratio
3
2
1
0
Significant (Yates)
Not-significant
Upper control limit
Lower control limit
significance boundary
-1
-2
0.0
0.4
0.8
1.2
standard error
9
Analysis of Ulcer Data of Moerman
Logistic regression model
Regression analysis
Response variate:
Y
Binomial totals:
n
Distribution: Binomial
Link function:
Logit
Fitted terms: Constant + Trial + Treat
Accumulated analysis of deviance
Change
d.f.
+ Trial
30
+ Treat
1
+ Treat.Trial
30
Total
61
(c) Stephen Senn 2007
mean
deviance
116.627
170.605
34.622
321.853
deviance
3.888
170.605
1.154
5.276
deviance approx
ratio
chi pr
3.89 <.001
170.60 <.001
1.15
0.257
10
Lessons from Moerman
•
There is no evidence of variation in the treatment
effect from trial to trial
We should be wary about concluding that apparent
variation signals true variation
We need to be cautious and think carefully about
analysis
Of course…it is always possible that there was
exactly the same genetic mix in each trial
•
•
•
–
•
in which case gene by treatment would not manifest itself
as trial by treatment interaction
We need to understand components of variation
(c) Stephen Senn 2007
11
What you learn in your first ANOVA
course
• Completely randomised design
– One way ANOVA
• Randomised blocks design
– Two way ANOVA
• Randomised blocks design with replication
– Two way ANOVA with interaction
• No replication, no interaction
(c) Stephen Senn 2007
12
1.
Senn SJ. Individual Therapy: New Dawn or False Dawn. Drug
Information Journal 2001;35(4):1479-1494.
(c) Stephen Senn 2007
13
(c) Stephen Senn 2007
14
Second cross-over
Responders
First
crossover
(c) Stephen Senn 2007
NonResponders
Total
Responders 24
0
24
Non0
Responders
8
8
Total
8
32
24
15
Second cross-over
Responders
NonResponders
Total
6
24
Non6
Responders
2
8
Total
8
32
Responders 18
First
cross-over
(c) Stephen Senn 2007
24
16
But Suppose you Only Have one
Cross-over
Second cross-over
First
Responders
cross-over
NonResponders
Total
(c) Stephen Senn 2007
Responders
NonResponders
Total
?
?
24
?
?
8
32
17
Two Strategies
Gene led
• Identify suitable loci
using in vitro studies
• Generate possible
treatment hypotheses
• Select suitable
patients
– ‘Enrichment’ studies
• Prove that the
treatment works for
these patients
(c) Stephen Senn 2007
Treatment led
• Identify potential
treatments
• Find those that work
in general
• Find those where
patient by treatment
interaction is
considerable
• Search for genetic
subgroups
18
Strategy 1 (Treatment led)
Whole genome matching
“Drug responses are not persistent affairs; they are
temporary characteristics. One therefore may ask
whether twin studies are necessary to assess the
genetic element in pharmacological responsiveness.To
measure the genetic component contributing to their
variability, it seems logical to investigate the response
variation by repeated drug administration to given
individuals, and to compare the variability of the
responses within and between individuals.”
Kalow et al, Pharmacogenetics,8, 283-289, 1998.
(c) Stephen Senn 2007
19
Physicians like within patient
studies but statisticians get cross
over them
The Sayings of Confuseus
(c) Stephen Senn 2007
20
Possible Strategy
• Run multi-period cross-overs
• Patient by treatment interaction becomes
identifiable
• This provides an upper bound for gene by
treatment interaction
– Because patients differ by more than their
genes
(c) Stephen Senn 2007
21
Advantages and Disadvantages
PRO
• Cheap
• Low tech
• Insight into sources
of variation gained
• Good at identifying
if there are gene by
treatment
interactions
(c) Stephen Senn 2007
CON
• Only suitable for
chronic diseases
• Demanding of
patient’s time
• Unglamorous
• Bad at identifying
which genes are
responsible for
treatment interactions
22
In Practice
• We hardly ever run repeated cross-over designs
• Hence we are incapable of telling formally which of the
two cases applies
• Most researchers simply assume by default that case 1
is the case that applies
• They assume that variation in response is a permanent
feature of patients
• This is what might be called patient-by-treatment
interaction and provides an upper bound for gene-bytreatment interaction
• Strangely enough, an area in which such repeated
cross-overs have been applied is one in which
interaction is unlikely to be important: bioequivalence
(c) Stephen Senn 2007
23
Shumaker and Metzler
“A single dose (125 mg), two-formulation four-period,
bioequivalence trial of phenytoin compared the test
product with the reference product. The study used the
replicated design:
RT T R
TR R T
where R is the reference product and T is the test
product. This design can be considered two replications:
Replicate 1
Replicate 2
RT
and
TR
Drug Information Journal,
Vol. 32, pp. 1063–1072,
TR
RT.”
1998
(c) Stephen Senn 2007
24
P
h
e
n
y
t
o
i
n
D
a
t
a
:
A
U
C
b
y
S
u
b
j
e
c
t
a
n
d
F
o
r
m
u
l
a
t
i
o
n
1
0
0
8
0
AUC
R
E
F
T
E
S
T
6
0
4
0
0
5
1
0
1
5
2
0
2
5
V
o
l
u
n
t
e
e
r
(c) Stephen Senn 2007
25
1
8
1
.
2
5
1
.
1
1
.
0
4
6
1
6
2
3
7
1
4
2
2
1
2
2
1
1
1
2
0
1
3
1
9
1
0 1
1
5
3
1
7 9
8
2
Relativbolaity:secondtrmiaon
0
.
9
2
5
2
6
2
4
0
.
8
0
.
8
0
.
9
1
.
0
1
.
1
1
.
2
R
e
l
a
t
i
v
e
b
i
o
a
v
a
i
l
a
b
i
l
i
t
y
f
i
r
s
t
d
e
t
e
r
m
i
n
a
t
i
o
n
(c) Stephen Senn 2007
26
Simple approach ignoring period
Accumulated analysis of variance
Change
d.f.
+ SUB
25
+ PROD
1
+ SUB.PROD
Residual
Total
s.s.
m.s.
v.r.
7.748
0.310
82.3
0.00253
0.00253
0.67
0.416
0.0679
0.00272
0.72
0.811
52
0.196
0.00377
103
8.014
0.0778
25
F pr.
<.001
Estimated variance components
Random term
component
SUB
0.076800
SUB.PROD
-0.000524
(c) Stephen Senn 2007
s.e.
0.021915
0.000533
27
Pharmacogenomics:
A subject with great promise.
(c) Stephen Senn 2007
28
Strategy Two (Gene Led)
Genetic Subgroups
• In many indications cross-over trials are
impossible
• This means that we have to investigate
interaction not by whole genome matching (each
patient his or her own control) but by genetic
subgroups
• Patients provide replication of the subgroup
– Which genes should we use?
– How should we group genotypes?
– Will we have the statistical power to investigate
subgroup interactions?
(c) Stephen Senn 2007
29

A Dose-Response
View of Genetics


Y X  EC50   
X


X  EC50
1
Phenotype
1
0.5
0
1
2
Allele copies
Dominant
Recessive
Additive
(c) Stephen Senn 2007
30
Pairs of Orthogonal Contrasts
Genotype
AA
Aa
aa
Score
0
1
2
Linear
-1
0
1
2
Quadratic
-1
2
-1
6
Dominant
-2
1
1
6
Within a
0
-1
1
2
Recessive
-1
-1
2
6
Within A
-1
1
0
2
(c) Stephen Senn 2007
Variance multiplier
See also Balding DJ Nat Rev Genet 2006;7(10):781-91.
31
One t-test versus one 2 DF F-test
 1.96
4
1.96
Quadratic contrast
2
4
2
0
2
4
2
4
Linear contrast
(c) Stephen Senn 2007
Second approach either the linear or quadratic approach is tested
32
Two t-tests versus one 2 DF F-test
 2.236
4
2.236
2.236
Quadratic contrast
2
4
2
0
2
2
4
 2.236
4
Linear contrast
(c) Stephen Senn 2007
33
Impact on trial design
• Suppose that you know that a dominant (with a as
dominant allele) model applies
• Then optimal clinical trial design implies that you should
have half the patients on AA and the other half on Aa or
aa
• But if HW equilibrium applies this will only happen
naturally if the probability of allele A is √2
• Of course, since disease is a selection process HW
equilibrium may not apply anyway but this does not get
around the problem
• The distribution of genotypes may be very unfavourable
for efficient investigation
(c) Stephen Senn 2007
34
Genotype frequency for Hardy-Weinberg equilibrium
Probability of genotype
1
0.5
0
0
0.2
0.4
0.6
0.8
1
Probability of allele a
AA
Aa
aa
Total
(c) Stephen Senn 2007
35
Contrast multipliers for three genotypes
Genotype multiplier
0.5
1
1
0
0.2
0.4
0.6
0.8
1
1
1
Probability of allele a
Aa
AA
aa
(c) Stephen Senn 2007
36
Variances for gene-by-treatment contrasts
1
1
1
2
Variance of contrast
2
4

0.2
0.4
2
N
0.6
0.8
Allele relative frequency
Linear
Dominant
Recessive
Universal
(c) Stephen Senn 2007
37
‘Enrichment’ studies?
• Could we fix enrollment so that we have
optimal genotype frequencies?
• Problems
– Recruitment time increases
– Only optimal for one given locus
– Requires knowledge of allele copy response
• Dominant, recessive, linear etc
– Requires knowledge of relevant locus
– Interferes with other purposes of trial
(c) Stephen Senn 2007
38
Pharmacoeconomics and
genotyping
• Finding a subset of patients who benefit has the
potential to make the market smaller
• This might imply that it is not in the economic
interests of sponsors to do so
• In fact models can be produced that suggest
subsetting is valuable
• An adaptation of a model of Kwerel(1980), which
was originally applied to another situation, will
be considered
(c) Stephen Senn 2007
39
Economic Model
  probability side effect, L  loss to patient
b  benefit, p  price, c  cost of sale
1
f b  , 0  b   ,    L  p


L p
db  1 
 proportion benefitting


 L p 
1
 L p
1 
  p  c   marginal revenue per patient
 

Crucial assumption: the sponsor can change the price
(c) Stephen Senn 2007
40
Pharmacogentic model
  probability low risk
1    probability low risk
 1  probability side effect given low risk
 2  probability side effect given low risk
 = 1  1     2
Suppose   0.86,  1  0.05,  2  0.3
Position is shown on next slide
(c) Stephen Senn 2007
41
Marginal revenue
0.1
0.05
0
0.2
0.4
0.6
0.8
Price
Perceived average risk
Low risk market
High risk market
Genotyped market
(c) Stephen Senn 2007
42
0.52
0.55
Marginal revenue
0.108
0.106
0.1055
0.104
0.1025
0.102
0.1
0.45
0.5
0.55
Price
Perceived average risk
Low risk market
High risk market
Genotyped market
(c) Stephen Senn 2007
43
An Issue with Covariates
• Covariate adjustment in clinical trials is generally beneficial and to
be recommended
– However a point to note is that the covariates in question should be
measured prior to allocation of treatment
– Otherwise problems arise with causal inference
– Some of the treatment effect may be removed
• However, when looking at gene-by-treatment interaction there is a
potential problem
• Covariates can be pre treatment allocation and hence unaffected by
treatment but can be affected by genetics
• Hence fitting the covariate could remove some of the gene effect
• Will inference about gene-by-treatment interaction still be sound?
• This issue requires careful thought
(c) Stephen Senn 2007
44
An Overlooked Source of Genetic
Variability
• Humans may be classified into two important
genetic subtypes
• One of these suffers from a massive
chromosomal deficiency
• This is expressed in
– important phenotypic differences
– a huge disadvantage in life expectancy
• Many treatment strategies take no account of this
• The names of these subtypes are...
(c) Stephen Senn 2007
45
Males and females
(c) Stephen Senn 2007
46