Transcript Document

The Impact of Epidemiology
on Advancing Genomic
Technologies
Muin J. Khoury, MD, Ph.D.
Director,
CDC Office of Genomics and Disease Prevention
Outline






Genomics 2005: Science and Expectations
Why do we need epidemiology?
HuGE: Characterizing genes in populations
Epidemiologic assessment of genomic tests
Epidemiologic assessment of family history as a
tool for disease prevention
Human Genome Epidemiology Network (HuGE
Net)
“Welcome to the Genomic Era”
Guttmacher and Collins, NEJM 2003;349:996

DNA 50th Anniversary

Human Genome Sequence
“DNA Changed the World: Now What?”
NY Times, February 25, 2003
How Will Genetics Change Our Lives
50 Years from Now?

“We will have individualized, preventive medical
care based on our own predicted risk of disease
as assessed by looking at our DNA. By then
each of us will have had our genomes
sequenced because it will cost less than $100 to
do that. And this information will be part of our
medical record. Because we will still get sick,
we'll still need drugs, but these will be tailored to
our individual needs.

F. Collins MD, PhD, TIME, the Future of Life, 2003
Predicted Home Computers for 2004 !?!
Popular Mechanics Magazine (1954)
From Genetics to Genomics
Genetics
Genomics

Disease

Information

Single Gene Disorders

All Diseases

Mutations/One Gene

Variation/Multi Genes

Inherited

Inherited/somatic

High Disease Risk

Low Disease Risk

Environment Role +/-

Environment Role ++

“Genetic Services”

General Practice
Genetic Disorders Featuring CAD/MI
Apolipoprotein(a) excess
Apolipoprotein AI deficiency
Autosomal recessive
hypercholesterolemia
Cerebrotendinous
xanthomatosis
Fabry disease
Familial combined
hyperlipidemia
Familial defective apoB
Familial
hypercholesterolemia
Familial partial lipodystrophy
Familial pseudo hyper
kalemia due to RBCl leak
Heparin cofactor II deficiency
Homocystinuria/homocysteinemi
a
Niemann-Pick disease, type E
Progeria
Protein C deficiency
Pseudoxanthoma elasticum
Sitosterolemia
Spontaneous coronary dissection
Tangier disease
Type III hyperlipoproteinemia
Werner syndrome
Williams syndrome
Gene-Environment Interaction in
Cardiovascular Disease

“Some vegetarians with
'acceptable' cholesterol
levels suffer myocardial
infarction in the 30's.
Other individuals...seem
to live forever despite
personal stress,
smoking, obesity, and
poor adherence to a
Heart Associationapproved diet"
R.A. Hegele (1992)
Genetics and Cardiovascular Disease
Stress
Nutrition
Homocysteine
Blood
Pressure
Diabetes
LDL
Cholesterol
Smoking
Health
Status
GENES
Obesity
Fibrinogen
Exercise
Triglycerides
Lp(a)
Insulin
Prediction of Risk of Myocardial Infarction from
Polymorphisms in Candidate Genes
Yamada et al. NEJM 2002;347:1916-1923.




Case-Control Study (5061 MI and 2242
Controls)
Analysis of 71 candidate genes with 112
polymorphisms (2-step process)
A few associations were found…small odds
ratios…
Accompanying editorial


“Findings should be used to initiate further research
Recommendations for primary prevention cannot be
based on these findings.”
Outline






Genomics 2005: Science and Expectations
Why do we need epidemiology?
HuGE: Characterizing genes in populations
Epidemiologic assessment of genomic tests
Epidemiologic assessment of family history as a
tool for disease prevention
Human Genome Epidemiology Network (HuGE
Net)
Gene-Based Medicine in 2010?
“Hypothetical Genetic Test Report”

Condition

Prostate Ca
HPC1, 2, 3
Alzheimer’s
APOE,FAD3,XAD
Heart disease APOB,CETP
Colon Cancer FCC4,APC
Lung Cancer
NAT2




Genes
RR
Lifetime
0.4
0.3
2.5
4.0
6.0
7%
10%
70%
23%
40%
Collins FC, New Engl J Med 1999;341:28-37.
Epidemiology in the 21st Century: Population
Impact of Human Genome Variation
“Calculation, Communication, and Intervention”
J Koplan (CDC Director, 2000)

“The sequencing of the
human genome offers the
greatest opportunity for
epidemiology since John
Snow discovered the
Broad Street pump”

Shpilberg et al. J Clin
Epidemiol (1997)
The Importance of Epidemiologic Data

Prevalence

Disease burden




Relative risk
Absolute risk
Attributable fraction
Gene-Environment
interaction
The Importance of Epidemiologic
Methods

Selection bias

Confounding bias

Information bias

Statistical power

Generalizability
Epidemiological Quality in Molecular Genetic
Research: Need for Methodological Standards




Bogardus et al. JAMA
1999;281:1919-26.
Aim: To examine clinical
epi quality of recent
papers in molecular
genetic analysis
40 inferential articles in 4
clinical journals (1995)
38% articles passed 4 or
less methodologic
standards

Methodologic
Standard

Reproducibility
38%
Objectivity
33%
Case Group
78%
Case Spectrum
88%
Comparison Group 70%
Comparison Spectrum
88%
Quantitative Summary
90%






%
The Importance of Population-based Data:
Risk of breast cancer in women with BRCA1/2
mutations
85% (‘94)

High risk families - (BRCA1/2)

56% (‘97)

Voluntary survey, Wash DC (BRCA1/2)

40% (‘99)

Population-based sample,
Australia - (BRCA1/2)

37% (‘98)

Population-based sample,
Iceland - (BRCA2)

Courtesy: Dr Wylie Burke
“Systematic application of epidemiologic
method and approaches to assess the
impact of human genetic variation on
health and disease”
Khoury, Little and Burke, HuGE 2004
• Genotype prevalence
• Gene - disease association
• Gene - gene interactions
• Gene - environment interactions
• Assessment of Genetic tests
HuGE problem:
30,000 genes, their
combinations and
interactions with risk
factors
The End of Black Box Epidemiology?
Risk Factors
Demographics
Diet
Occupation
Smoking
Alcohol
Environment
Adverse
Health
Outcomes
Inflammation
Activation of
Maternal/Fetal HPA
Axis
Decidual
Hemorrhage
Abruption
• Infection:
- Chorion-Decidual
- Systemic
• Maternal-Fetal
Prothrombin G20210A
Factor V Leiden
Protein C, Protein S
Type 1 Plasminogen
MTHFR
Stress
•Premature
Onset of
Physiologic
Initiators
CRH
E1-E3
CYP1A1
GSTT1
Susceptibility
to
environmental
toxins
proteases
Pathological
Uterine
Distention
Interleukin
s
TNF-α
Fas L
Gap jct
IL-8
PGE2
Oxytocin recep
Chorion
Decidua
+
+
• Multifetal
Pregnancy
• Polyhydramnios
• Uterine
abnormalities
Mechanical stretch
CRH
MMPs
uterotonins
PROM
Cervical change
Courtesy S. Dolan MD
Preterm
Delivery
Uterine
Contractions
Adapted from: C. J. Lockwood, Paediatr Perinat Epidemiol 15, 78 (2001) & X. Wangl. Paediatr Perinat Epidemiol 15, 63 (2001)
Outline






Genomics 2005: Science and Expectations
Why do we need epidemiology?
HuGE: Characterizing genes in populations
Epidemiologic assessment of genomic tests
Epidemiologic assessment of family history as a
tool for disease prevention
Human Genome Epidemiology Network (HuGE
Net)
Examples of Epidemiologic Efforts to Characterize
Gene Effects on Population Health

Cohort Studies

Biobanks

Case-Control Studies

National Birth Defects Study

Acute Public Health
Investigations

Leptospirosis outbreak

Cross-Sectional Studies

NHANES
Centers for Birth Defects
Research and Prevention
Arkansas
California
Georgia
Iowa
Massachusetts
New Jersey
New York
North Carolina
Texas
Utah
National Birth Defects Prevention Study

Case-control study of 24 major birth
defects started in 1997

Based on state surveillance systems

Maternal interview

Buccal cell (DNA) collection from infant
and parents
Studying Gene-Environment
Interaction
Smoking, detoxification genes, and orofacial clefts
Questions from interview:
• Did you ever smoke cigarettes?
• At any time from 2/3/97 to
1/28/98 did you smoke?
• During which months?
• How many cigarettes?
Baby with cleft lip
DNA for examination of
genes involved in
detoxification
Is there
gene –
environment
interaction?
Integrating Human Genomics into
Acute Public Health Investigations
Environment
Susceptibility
genes
Outbreaks
1998 Springfield Ironhorse Triathlon
Leptospirosis Outbreak
•
876 triathletes; 12% reported illness
•
Serum from 474; 52 positive for leptospirosis
Genetic studies: TNF-a, HLA-DRB, HLA-DQB
•
DNA from 85 anonymized blood samples
•
HLA-DQ6 positive triathletes (compared to DQ6 negatives)
were
- more likely be seropositive for leptospirosis
(OR=2.8, p=0.04)
- especially for those who reported swallowing lake water
(OR=8.5, p=0.001)
Lingappa J. et al., Genes & Immunity 2004;5:197-202
Outline






Genomics 2005: Science and Expectations
Why do we need epidemiology?
HuGE: Characterizing genes in populations
Epidemiologic assessment of genomic tests
Epidemiologic assessment of family history as a
tool for disease prevention
Human Genome Epidemiology Network (HuGE
Net)
”Genetic Testing for Sale”
Vineis, P, Christiani, D. Epidemiology Jan 2004
AmpliChip CYP450 Microarray ASR



2D6 and 2C19 genes - “…role in
metabolism of ~25% of prescription
drugs”
Chip = microarray detection system to
identify 29 2D6 and 2 2C19 alleles
Variations affect how common drugs
are processed or metabolized



Poor metabolizers  adverse
reactions
Ultrarapid metabolizers  nonresponders
“….aid for physicians in individualizing
treatment doses for patients on
therapeutics metabolized through
these genes.”
FDA clearance (510K) for
diagnostic use Dec, 2004
Clearance in EU
Diagnosing ovarian cancer by proteomics
Patterns of specific serum
proteins can be used to
detect OvCa, even in early
stages
 Clinical trials in progress
 FDA review will follow
 “OvaCheck” technology &
interpretive software
licensed
 Scheduled to be offered in
2004 but many questions

Baggerly et al and Ransohoff
JNCI, Feb 2005
Petricoin EF. Use of proteomic
patterns in serum to
identify ovarian cancer. Lancet. 2002
Feb 16;359(9306):572-7.
Public Health Approach to Evaluation of
Genetic Tests (ACCE Model System)

Effective
Intervention
(Benefit)
Disorder/Setting
Natural
History




Analytic Validity
Clinical Validity
Clinical Utility
ELSI
Clinical
Specificity
Ethical, Legal, &
Social Implications
(safeguards& impediments)
Quality
Assurance
Clinical
Sensitivity Prevalence
Pilot
Trials
PPV
NPV
Disorder
&
Setting
Penetrance
Analytic
Assay
Sensitivity
Robustness
Analytic Quality
Specificity Control
Monitoring
&
Evaluation
Education
Facilities
Health
Risks
Economic
Evaluation

Analytic Validity
Defines the
ability of a test to
accurately and
reliably identify
or measure the
analyte or
mutation of
interest
Effective
Intervention
(Benefit)
Natural
History
Clinical
Specificity
Ethical, Legal, &
Social Implications
(safeguards& impediments)
Quality
Assurance
Clinical
Sensitivity Prevalence
Pilot
Trials
PPV
NPV
Disorder
&
Setting
Penetrance
Analytic
Assay
Sensitivity
Robustness
Analytic Quality
Specificity Control
Monitoring
&
Evaluation
Education
Facilities
Health
Risks
Economic
Evaluation
Analytic sensitivity & specificity
Sensitivity: Proportion of positive results
when variant/analyte is present
Specificity: Proportion of negative results
when variant/analyte is absent
• Measures intrinsic performance of assay technology
• Part of laboratory validation before use
• Established using positive and negative control
samples characterized using “gold standard” or by
consensus


Clinical validity
Defines the ability
of a test to detect
or predict the
phenotype or
particular clinical
outcome
Elements build
upon analysis of
analytic validity
Effective
Intervention
(Benefit)
Natural
History
Clinical
Specificity
Ethical, Legal, &
Social Implications
(safeguards& impediments)
Quality
Assurance
Clinical
Sensitivity Prevalence
Pilot
Trials
PPV
NPV
Disorder
&
Setting
Penetrance
Analytic
Assay
Sensitivity
Robustness
Analytic Quality
Specificity Control
Monitoring
&
Evaluation
Education
Facilities
Health
Risks
Economic
Evaluation
Clinical sensitivity & specificity
Test Result
Disease
Phenotype
Yes
No
Pos
A
B
Neg C
D
Sensitivity: Proportion of positive test results in
individuals who have the phenotype = A / (A+C)
Specificity: Proportion of negative test results in
individuals who do not have the phenotype = D / (B+D)
Positive & negative predictive values
Disease
Phenotype
Yes
No
Test Result
Pos
Neg
A
C
B
D
Positive predictive value = A / (A+B)
Probability that person with positive test will have the
phenotype
Negative predictive value = D / (C+D)
Probability that person with negative test will not
have the phenotype
Effect of prevalence on Clinical PPV
50
45
PPV = 45%
99,950
100
NPV = 99.9%
100,000
Prevalence 5 in 10,000, Sens = 90%, Spec = 99.9%
500
450
PPV = 82%
99,500
100
NPV = 99.9%
100,000
Prevalence 50 in 10,000, Sens = 90%, Spec = 99.9%


Clinical Utility
Defining the risks
and benefits
associated with
introduction into
routine clinical
practice
Likelihood of
improved health
outcome
Effective
InterventionOutcomes
Natural
History
Clinical
Specificity
Ethical, Legal, &
Social Implications
(safeguards& impediments)
Quality
Assurance
Clinical
Sensitivity Prevalence
Pilot
Trials
PPV
NPV
Disorder
&
Setting
Penetrance
Analytic
Assay
Sensitivity
Robustness
Analytic Quality
Specificity Control
Monitoring
&
Evaluation
Education
Facilities
Health
Risks
Economic
Evaluation
Case Studies of Clinical Utility:
Using Epidemiologic Data for Policy

Should we screen the general population
for hereditary hemochromatosis?

Should we screen women for Factor V
Leiden before prescribing oral
contraceptives?

Should we screen children for TPMT
deficiency before ALL rx with 6MP?
Case Study 1:
Hereditary Hemochromatosis








“The Genetic Disorder of
the 21st Century”
Iron Overload
Multiple organ system
Intervention: simple
Gene Chromosome 6
1997 Expert Panel on
Population Screening
Public health research
Agenda
Provider education
campaign
Prevalence of
Hereditary Hemochromatosis
Mutations in the USA
NHANES III
Genotype Prevalence (%)
Genotype/Group
White
Black
C282Y/C282Y
H63D/H63D
C282Y/H63D
0.3
2.2
2.4
Hisp
.06
0.3
.06
Steinberg KK et al., JAMA 2001;285:2216
.03
1.1
0.2
Hemochromatosis-Associated Hospitalizations,
National Hospital Discharge Survey 1979-1997
Rate per 100,000 US residents
5
4
3
2
1
males
females
0
79 - 82
83 - 87
88 - 92
Years
Brown al et al. Genet Med 2001;3:109-111
93 – 97
Body Iron Content in grams
Natural History of Hereditary Hemochromatosis
30
Early death
25
Bronze diabetes
20
Signs of organ damage
15
Non-specific symptoms
10
Asymptomatic
5
Mutation
0
0
10
20
30
Age
40
50
60
Case Study 2:
Incidence of Venous Thrombosis Among Women
by Factor V Leiden and Oral Contraceptive Use
30
Incidence/10,000
25
20
15
10
5
0
(-) Mutation
(-) OC
(-) Mutation
(+) OC
(+) Mutation
(-) OC
(+) Mutation
(+) OC
Mutation = Factor V Leiden; OC = Oral Contraceptives
Source: Adapted from Vandenbroucke JP, et al. BMJ 1996; 313: 1127-1130
Screening for factor V Leiden mutation before
prescribing oral contraceptives?

Cost-effectiveness of screening for factor V Leiden
mutation in women in the United States

To prevent one venous thromboembolic death
attributable to oral contraceptives in women with factor V
Leiden mutation, >92,000 carriers need to be identified
and stopped from using these pills

Estimated charge to prevent this one death exceeds
$300 million

Creinin MD et al. Fertil Steril 1999;72(4):646-51
Case Study 3:
Genetic Testing (TPMT) Decision Analysis Tree for 6MP Therapy for Acute Lymphoblastic Leukemia
(Veenstra et al. AAPS Pharmasci 2000;2)
Influence of Cost of Genetic Test and Outcome
Severity on Hypothetical Cost-Effectiveness of
Genotyping (Veenstra et al)
Genotype Prevalence=0.5%
Influence of Cost of Genetic Test and Outcome
Severity on Hypothetical Cost-Effectiveness of
Genotyping (Veenstra et al)
Genotype Prevalence=1%
Influence of Cost of Genetic Test and Outcome
Severity on Hypothetical Cost-Effectiveness of
Genotyping (Veenstra)
Genotype Prevalence=0.3%
Framework for Evaluating the Potential CostEffectiveness of Pharmacogenomics
(Veenstra et al. AAPS Pharmasci 2000;2)

FACTORS

FACTORS for COSTEFFECTIVENESS

Outcome severity
Drug monitoring
Geno-Pheno Corr
Assay
Polymorphism

+++
NA/difficult
+++
Rapid, inexpensive
High allele frequency








Outline






Genomics 2005: Science and Expectations
Why do we need epidemiology?
Characterizing genes in populations
Epidemiologic assessment of genomic tests
Epidemiologic assessment of family history as a
tool for disease prevention
Human Genome Epidemiology Network (HuGE
Net)
Family history –
a “genomic tool”
•
Most common diseases - interactions of multiple
genes, behaviors, and environment
•
Relatives share a lot of genes in common (half
for 1st rel, quarter for 2nd rel, and so on)
•
A long way to go for genetic risk factor testing
•
Family medical history is a “genomic tool” that
captures these interactions
Family History of Common Diseases
57%
No family history
33%
One disorder
2%
Three or more disorders
8%
Two disorders
Scheuner et al. Am J Med Genet 1997;71:315-324.
Family history is a risk factor
for almost all common diseases
Relative Risk
Heart disease
Breast cancer
Colorectal cancer
Prostate cancer
Melanoma
Type II diabetes
Osteoporosis
Asthma
2.0 – 5.4
2.1 – 3.9
1.7 – 4.9
3.2 – 11.0
2.7 – 4.3
2.4 – 4.0
2.0 – 2.4
3.0 – 7.0
Am J Prev Med - February 2003
Family History
Public Health Initiative
Why focus on family history?
•
FHx is underutilized in preventive medicine
•
Risk factor for most chronic diseases of PH significance
How can we use family history?
•
assess risk for common chronic diseases
•
influence early screening for disease
•
educate people about prevention measures
Family History Collection by PCPs

Family history collected at about 50% of new visits
and 22% of established visits

Average duration of visit, 10 minutes; average
duration of family history discussion, 2.5 minutes
Acheson et al., 2000

Only 29% of PCPs feel prepared to take family
history and draw pedigrees
Suchard et al., 1999
The Health Family Tree Study
Utah, 1983-1996 - CHD
FHx Score
0.5 (positive)
1.0 (str pos)
2.0 (v str pos)
% Families
% Early CHD
% All CHD
14
3.2
1.0
72.1
34.7
16.8
48.4
17.6
6.3
Includes data from 122,155 families; 16,602 early CHD cases;
54,182 cases of CHD at any age
Williams, et al. Am J Cardiol 2001; 87:129-135
Schema for Using Family History to
Guide and Inform Prevention Activities
Average
Family
History
Tool
Moderate
Standard prevention
recommendations
Personalized
prevention
recommendations
High
Genetic Evaluation
+ personalized
prevention
recommendations
Outline






Genomics 2005: Science and Expectations
Why do we need epidemiology?
Characterizing genes in populations
Epidemiologic assessment of genomic tests
Epidemiologic assessment of family history as a
tool for disease prevention
Human Genome Epidemiology Network (HuGE
Net)
Human Genome Epidemiology
Network (HuGE Net)

Global collaboration
of individuals and
organizations to
assess population
impact of genomics
and how it can be
used to improve
health and prevent
disease




Information Exchange
Training and Technical
Assistance
Knowledge Base
Development
Information
Dissemination
From HuGE Research to Synthesis &
Dissemination for Policy and Practice
Primary HuGE Research
Agenda and Funding
Study Design
Single studies
(case-control,
Implementation
cohort
Candidate gene selection
biobanks)
Risk factor data
Consortia
Outcome data
Identify
Knowledge gaps
Methodologic problems
Research Priorities
Analysis: G-G, G-E
Interpretation
Causal inference
Risk estimation
Synthesis
Policy &
Practice
Reviews
Meta-analysis
Appraisal
(Single study)
Reporting
From HuGE Research to Synthesis &
Dissemination for Policy and Practice
Primary HuGE Research
Agenda and Funding
Study Design
Single studies
(case-control,
Implementation
cohort
Candidate gene selection
biobanks)
Risk factor data
Consortia
Outcome data
Identify
Knowledge gaps
Methodologic problems
Research Priorities
HuGE Net
Analysis: G-G, G-E
Interpretation
Causal inference
Risk estimation
Synthesis
Policy &
Practice
(EGAPP)
Reviews
Meta-analysis
Appraisal
(Single study)
Reporting
The Human Genome Epidemiology Network
Research agenda
Study design
Implementation
•
Analysis
Research priorities
Interpretation
Policy
Practice
Synthesis
Appraisal
Dissemination
Selected HuGE Net Activities







HuGE Studies Database
HuGE Reviews
Methodology/Training Workshops
International Biobank/Cohort Study
meeting
“Network of Networks”
HuGE book
HuGE journal
No. of articles in Huge Published
Literature db, 2001-2004*
4500
4079
4000
3504
3500
3160
# of articles
3000
2462
2500
2000
# Pub Lit articles
1500
1000
500
0
2001
2002
2003
2004
Year of publication
(*As of Jan 14, 2005. Count excludes review articles, m eta & pooled analyses.)
Number of Published HuGE Papers*
2001-2004

Year
Prevalence Associations Interactions

2001
308
2141
436

2002
349
2799
569

2003
323
3010
598

2004
368
3486
604
Rank
1
Gene
Symbol
Gene name
APOE
apolipoprotein E
481
ACE
angiotensin I converting enzyme (peptidyldipeptidase A) 1
398
MTHFR
5,10-methylenetetrahydrofolate reductase
(NADPH)
377
HLADRB1
major histocompatibility complex, class II, DR
beta 1
376
TNF
tumor necrosis factor (TNF superfamily, member
2)
346
2
3
4
5
# of
Papers
01-03
6
GSTM1 glutathione S-transferase M1
253
major histocompatibility complex, class II, DQ
beta 1
248
7
HLADQB1
8
F5
coagulation factor V (proaccelerin, labile factor)
213
GSTT1 glutathione S-transferase theta 1
9
204
The Need for Integrating Epidemiologic Evidence:
on Genes and Population Health







Unmanageable amounts of data
Small sample size of individual studies
Small effect size of gene-disease associations
Replication of associations
Publication bias
Heterogeneity
Generate and test hypotheses
Small sample size of individual studies
160
140
120
100
80
60
40
20
0
0.
85
0
0.
65
0
0.
45
0
0.
25
.0
50
.0
50
18
.0
50
16
.0
50
14
.0
50
12
.0
50
10
0
Sample size
Ioannidis, Trends Mol Med 2003
Small effect sizes in individual studies
120
100
80
60
40
20
0
0.0
.2
.4
Odds ratio
.6
.8
1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8
Outline






Genomics 2005: Science and Expectations
Why do we need epidemiology?
Characterizing genes in populations
Epidemiologic assessment of genomic tests
Epidemiologic assessment of family history as a
tool for disease prevention
Human Genome Epidemiology Network (HuGE
Net)