Upload
stephanie-simmons
View
220
Download
3
Tags:
Embed Size (px)
Citation preview
(c) Stephen Senn 2007 1
Pharmacogenetics - difficult or just impossible?
Stephen Senn
(c) Stephen Senn 2007 2
Based on chapter 25 (with some additional material from chapter 24).
(c) Stephen Senn 2007 3
“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-the-peg suit and one made to measure.
Chris Harbron, Significance, June 2006, p67 (My italics)
(c) Stephen Senn 2007 4
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 5
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 6
Pharmacogenetics: A cutting-edge science that will start delivering miracle cures the year after next.
(c) Stephen Senn 2007 7
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 8
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 9
0.0 0.4 0.8 1.2
standard error
-2
-1
0
1
2
3
4
log-
odds
rat
io31 Placebo-Controlled Trials of Cimetidine
Significant (Yates)Not-significantUpper control limitLower control limitsignificance boundary
(c) Stephen Senn 2007 10
Analysis of Ulcer Data of Moerman Logistic regression model Regression analysis
Response variate: YBinomial totals: nDistribution: BinomialLink function: LogitFitted terms: Constant + Trial + Treat
Accumulated analysis of deviance
mean deviance approxChange d.f. deviance deviance ratio chi pr+ Trial 30 116.627 3.888 3.89 <.001+ Treat 1 170.605 170.605 170.60 <.001 + Treat.Trial 30 34.622 1.154 1.15 0.257Total 61 321.853 5.276
(c) Stephen Senn 2007 11
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 12
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 13
1. Senn SJ. Individual Therapy: New Dawn or False Dawn. Drug Information Journal 2001;35(4):1479-1494.
(c) Stephen Senn 2007 14
(c) Stephen Senn 2007 15
Second cross-over
Responders Non-Responders
Total
First cross-over
Responders 24 0 24
Non-Responders
0 8 8
Total 24 8 32
(c) Stephen Senn 2007 16
Second cross-over
Responders Non-Responders
Total
First cross-over
Responders 18 6 24
Non-Responders
6 2 8
Total 24 8 32
(c) Stephen Senn 2007 17
But Suppose you Only Have one Cross-over
Second cross-over
Responders Non-Responders
Total
First cross-over
Responders ? ? 24
Non-Responders
? ? 8
Total 32
(c) Stephen Senn 2007 18
Two StrategiesGene led Treatment 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
• Identify potential treatments
• Find those that work in general
• Find those where patient by treatment interaction is considerable
• Search for genetic subgroups
(c) Stephen Senn 2007 19
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 20
Physicians like within patient studies but statisticians get cross over them
The Sayings of Confuseus
(c) Stephen Senn 2007 21
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 22
Advantages and DisadvantagesPRO CON
• Cheap• Low tech• Insight into sources
of variation gained• Good at identifying
if there are gene by treatment interactions
• Only suitable for chronic diseases
• Demanding of patient’s time
• Unglamorous• Bad at identifying
which genes are responsible for treatment interactions
(c) Stephen Senn 2007 23
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-by-treatment 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 24
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 TR RT.”
Drug Information Journal, Vol. 32, pp. 1063–1072, 1998
(c) Stephen Senn 2007 25
0 5 10 15 20 25
Volunteer
40
60
80
100
AU
CPhenytoin Data: AUC by Subject and Formulation
REFTEST
(c) Stephen Senn 2007 26
0.8 0.9 1.0 1.1 1.2
Relative bioavailability first determination
0.8
0.9
1.0
1.1
1.2
Rel
ativ
e bi
oava
ilabi
lity:
sec
ond
dete
rmin
atio
n
1
23
4
5
67
89
1011
1213
14
15
16
17
18
19202122 23
242526
(c) Stephen Senn 2007 27
Simple approach ignoring period
Accumulated analysis of variance
Change d.f. s.s. m.s. v.r. F pr.
+ SUB 25 7.748 0.310 82.3 <.001
+ PROD 1 0.00253 0.00253 0.67 0.416
+ SUB.PROD 25 0.0679 0.00272 0.72 0.811
Residual 52 0.196 0.00377
Total 103 8.014 0.0778
Estimated variance components
Random term component s.e.
SUB 0.076800 0.021915
SUB.PROD -0.000524 0.000533
(c) Stephen Senn 2007 28
Pharmacogenomics:
A subject with great promise.
(c) Stephen Senn 2007 29
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 30
A Dose-Response View of GeneticsY X EC50 X
X
EC50
0 1 2
0.5
1
DominantRecessiveAdditive
DominantRecessiveAdditive
Allele copies
Phe
noty
pe1
(c) Stephen Senn 2007 31
Pairs of Orthogonal Contrasts
Genotype AA Aa aa
Score 0 1 2 Variance multiplier
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
See also Balding DJ Nat Rev Genet 2006;7(10):781-91.
(c) Stephen Senn 2007 32
4 2 0 2 4
4
2
2
4
One t-test versus one 2 DF F-test
Linear contrast
Qua
drat
ic c
ontr
ast
1.96 1.96
Second approach either the linear or quadratic approach is tested
(c) Stephen Senn 2007 33
4 2 0 2 4
4
2
2
4
Two t-tests versus one 2 DF F-test
Linear contrast
Qua
drat
ic c
ontr
ast
2.236
2.236
2.236 2.236
(c) Stephen Senn 2007 34
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 35
0 0.2 0.4 0.6 0.8 10
0.5
1
AAAaaaTotal
Genotype frequency for Hardy-Weinberg equilibrium
Probability of allele a
Pro
babi
lity
of g
enot
ype
(c) Stephen Senn 2007 36
0 0.2 0.4 0.6 0.8 1
1
1
AaAAaa
Contrast multipliers for three genotypes
Probability of allele a
Gen
otyp
e m
ultip
lier
1
1
0.5
(c) Stephen Senn 2007 37
0.2 0.4 0.6 0.8
LinearDominantRecessiveUniversal
Variances for gene-by-treatment contrasts
Allele relative frequency
Var
ianc
e of
con
tras
t
4
11
2
1
2
N2
(c) Stephen Senn 2007 38
‘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 39
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 40
Economic Model
probability side effect, loss to patient
benefit, price, cost of sale
1,0 ,
11 proportion benefitting
1 marginal revenue per patient
L p
L
b p c
f b b L p
L pdb
L pp c
Crucial assumption: the sponsor can change the price
(c) Stephen Senn 2007 41
Pharmacogentic model
1
2
1 2
1 2
probability low risk
1 probability low risk
probability side effect given low risk
probability side effect given low risk
= 1
Suppose 0.86, 0.05, 0.3
Position is shown on next slide
(c) Stephen Senn 2007 42
0 0.2 0.4 0.6 0.8
0.05
0.1
Perceived average riskLow risk marketHigh risk marketGenotyped market
Price
Mar
gina
l rev
enue
(c) Stephen Senn 2007 43
0.45 0.5 0.550.1
0.102
0.104
0.106
0.108
Perceived average riskLow risk marketHigh risk marketGenotyped market
Price
Mar
gina
l rev
enue
0.1055
0.1025
0.52 0.55
(c) Stephen Senn 2007 44
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 45
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 46
Males and females