Transcript CCEB

Pharmacogenetics of
Leukemia Treatment
Response
Richard Aplenc
May 2nd, 2008
CCEB
Pediatric Leukemia
 Most common pediatric malignancy
 Four types




ALL
AML
CML
JMML
CCEB
Leukemia Treatment
 Varies both by disease and treating group
 Generally curable
 ~80% in ALL
 ~60% in AML
 Toxicity important
 Long term effects in ALL
 Infection and cardiac toxicity in AML
CCEB
Leukemia Treatment




Multi-agent
Over time
Substantial impact on patient and family
Accurate response prediction is clinically
very important
CCEB
ALL Therapy
Induction
L-Asp
Consolidation
Steroids
MTX
Interim
Maintenance
VCR
6-MP/6-TG
Delayed
Intensification
AraC
Maintenance
Doxorubicin Cyclophosphamide
CCEB
Predicting Treatment Response
 Leukemic blast characteristics
 Morphology
 Cytogenetics
 Molecular alterations (BCR-ABL)
 Patient characteristics
 Age
 Gender
 Genetic information?
CCEB
Genetic Information
 Variation in DNA sequence throughout the
genome
 Types of variation include
 Gene deletions (GSTT1)
 Duplications of DNA regions (TS 28 bp)
 Changes in single base pairs (SNPs)
 Allele, genotype, haplotype
CCEB
Allele/Genotype/Haplotype/CNV

SNP: Single Nucleotide Polymorphism

An allele is a single value for a single
marker

A genotype is a pair of alleles for a given
marker and both chromosomes in a
single person

Copy number variation (CNV) of DNA
sequences
Genotype
SNP 1
SNP 2
Haplotype
A haplotype is an ordered series of
alleles for many markers on a single
chromosome
Allele
...

Chromosome Chromosome from
from one parent
other parent
SNP example:
G
GTACGTTCG GGGCGGGAT
T
CCEB
Impact of Genetic Variability
 Loss of gene = loss of function
 Duplication of DNA segments and single
base pair changes may have different
effects depending on position
 Gain of function, loss of function, no change
CCEB
Our Dream
One Genotype Would Explain
Treatment Response
Why Did We Have This Dream?
 Thiopurine methylatransferase (TPMT)
 Low frequency variants have complete loss of
thiopurine metabolizing abilities
CCEB
That Dream Has Ended
Why Is That?
CCEB
CCEB
TPMT
One Gene, One Pathway, One Exposure
SH
TX
TU
HO
Allopurinol
N
N
SH
OH
N
N
SH
N
N
TPMT
SCH 3
N
N
N
HO
HGPRT
N
N
PO 4 CH 2
TIMP
TXMP
TGMP
O
H
N
N
N
XO
H
N
N
Mercaptopurine
HO
OH
TPMT Deficiency
H
6-MMP
CCEB
Two Remaining Questions
CCEB
Question 1:
Can we utilize data on host
genetic variability in a clinically
meaningful way?
Question 2:
Is Theo Zaoutis really Neo?
This Makes Sense Because…
Lisa Z looks like
Trinity
And Because…
Paul Offit is clearly
Morpheus
Now That Everyone is Awake…
 Return to Question 1
CCEB
Moving Towards the Answer
 Decide on the question
 Understand the complex phenotype issues
 Host genetics
 Environment
 Address the genetic epidemiology issues
CCEB
What is the Question?
 Does the genotype inform us of the
biology underlying a clinical outcome?
 Etiology
 Does the genotype predict a clinical
outcome?
 Prediction
CCEB
One Conceptual Approach
 Etiology
 Sensitivity
 Probability of positive test given disease
 Prediction
 Positive predictive value
 Probability of disease given positive test
 Seems obvious but impacts analysis
CCEB
Complex Phenotype: Host Genetics
 Common SNPs will have modest effects
 Potentially large impact for the population
 Rare SNPs may have bigger effects
 Small population impact
 SNP frequency and the effect size
determine sample size
 SNP frequency varies by ethnicity
CCEB
Complex Phenotype: Environment
 Identify and measure relevant covariates
 Genotype does not matter if the patient
doesn’t take the medication
 Concomitant medications
 Drug-drug interactions
 Alternative medications
 Folic acid supplimentation
 Other environmental exposures
CCEB
What are the Genetic Epidemiology
Issues?
 Population stratification
 Variation of SNP frequency by ethnicity
 High dimensional data
 Gene-environment interactions
 Interaction of host genetics with environment
 Gene-gene interactions
 Interaction of different SNPs
 Multiple comparisons
CCEB
Some Examples from Our Data
 Methotrexate interrupts the folate cycle
 ALL blasts are sensitive to folate depletion
 Polymorphisms in genes in the folate cycle
may impact methotrexate efficacy
CCEB
Relapse Free Survival by MTHFR C677T Variant Allele
1.00
0.90
0.80
0.70
0.60
p = 0.0486
0.50
0
5
10
Years
Wildtype (C)
Variant (T)
MTHFR C677T Cox Model
Covariate
HR
p
95% CI
C677T variant
1.93
0.004
1.229
3.037
Day 7 BM
1.77
0.013
1.125
2.773
Age
1.11
0.016
1.020
1.220
Race
1.71
0.307
0.610
4.798
Gender
1.37
0.238
0.811
2.323
Rx Arm
1.18
0.214
0.908
1.535
WBC
0.99
0.335
0.971
1.010
Phenotype
0.95
0.776
0.661
1.362
CCEB
CCEB
MTHFR C677T and Infection Risk
Gene
MTHFR
C677T
MTHFR
C677T
MTHFR
C677T
Genotype
N
C/C
C/T
T/T
C/C
C/T
T/T
C/C
C/T
T/T
224
187
72
224
187
72
224
187
72
Num.
Infection Type
Infection
46
Sepsis
42
Sepsis
16
Sepsis
155
Fever/Neutropenia
120
Fever/Neutropenia
53
Fever/Neutropenia
123
Infection - Other
113
Infection - Other
43
Infection - Other
OR
1
1.13
1.13
1
0.83
1.32
1
1.27
1.2
95% CI
P value
0.700-1.818
0.585-2.188
0.86
0.546-1.276
0.709-2.447
0.34
0.850-1.887
0.690-2.087
0.49
CCEB
MTHFR Conclusions
 The MTHFR C677T variant allele seems to
impact relapse risk
 Dose adjustment of methotrexate for
toxicity/infection does not ameliorate this effect
 Dose adjustment based on genotype may be
clinically useful
 Replication in anther sample set is ongoing
CCEB
MTFHR Issues
 Allele versus genotype versus haplotype
 Clinically meaningful analysis
 Positive predictive value
CCEB
Relapse Free Survival by MTHFR C677T Variant Allele
1.00
0.90
0.80
0.70
0.60
p = 0.0486
0.50
0
5
10
Years
Wildtype (C)
Variant (T)
Relapse Free Survival by MTHFR C677T Genotype
1.00
0.90
0.80
0.70
0.60
CC vs TT, p = 0.0477
0.50
0
5
10
Years
Wildtype (CC)
Variant (TT)
Heterozygote (CT)
Kaplan-Meier survival estimates, by haplo
1.00
0.75
0.50
0.25
0.00
0
5
10
analysis time
CA CA
CG CG
TA CG
TA TG
TG TG
CA CG
TA CA
TA TA
TG CG
15
PPV with Time to Relapse Data
 This is the metric of interest to oncologists
 Moscowitz and Pepe defined positive
predictive value in survival time data
 PPVXk(t) = P(T ≤ t | Xk = 1)
CCEB
PPV Conclusions
 Although statistically significant, the
MTHFR C677T allele has a PPV of 35%
 This is worse than flipping a coin
 Important question is the increased predictive
value above baseline
CCEB
TS 28 bp as Example
N
RFS
HR
CI
p
2R/2R
83
80%
1
--
--
2R/3R
196
79%
1.68
0.863-3.255
0.13
3R/3R
103
73%
1.87
0.942-3.721
0.074
3R/4R
20
60%
3.69
1.436-9.481
0.007
CCEB
TS 28 bp Bootstrapping
 Does knowledge of TS genotype improve
prediction of relapse?
 Bootstrap comparison of relapse free
survival of all patients with those with
particular TS polymorphisms
 No additional predictive value from
knowing TS genotype
 Caveat of sample size issues
CCEB
Other Genetic Epidemiology
Issues
 Multiple comparisons
 Gene-gene and gene-environment
interactions
CCEB
Multiple Comparisons
 Probability of finding a false association by
chance = 1 - 0.95n
 n = 10, p = 40%
 n = 100, p = 99.4%
 Our data:
 19 genotypes, 2 genders, 3 different relapse
sites
 N = 228, p = 99.99959%
CCEB
Methods for Multiple Comparisons
 Ignore it
 Validation sample set
 Adjust p-values
 Bonferroni
 False discovery rate (FDR) Benjamini et al 2001
 Use Bayesian methods
 False positive report probability (FPRP) Wacholder et al 2004
CCEB
High Dimensional Data
 The number of cells (N) needed to split R
variables into X partitions:
R
N=X
 A single 2-way combination
 R = 2, X= 3, N= 9
 We have evaluated 19 genotypes
 All 2-way combinations of our genotypes
 R = 19, X = 3, N = 1,162,261,467
CCEB
High Dimensional Data Methods
 Several methods in current use
 We have used patterning with recursive
partitioning (CART)
 Create groups as uniform as possible
 Use with genotype and other covariates
 No p-values
 Confirmation by cross-validation within the
sample set
CCEB
CCEB
CART Caveats
 No p-values
 Need to validate in a separate sample
 Often difficult to interpret results,
particularly of higher order interactions
 i.e. 2 genotypes and 1 environmental factor
CCEB
Future Directions
 Validate and extend genotyping in another
ALL sample set
 Incorporate drug dose data
 Investigate the impact of genetic variability
on infection risk in pediatric myeloid
leukemia
 R01 resubmission with Theo Zaoutis
CCEB
The End….
Thanks to everyone who makes it safe to swim with the sharks. Bev Lange,
Tim Rebbeck,Jinbo Chen, Theo Zaoutis, Tom McWilliams, Peggy Han,
Shannon Smith, Michelle Horn, Melanie Doran. Funded by RO1 CA108862-01.
CCEB