Plenary 3 Co-chairs: Sally Green and Helen Whelton Pembroke

The Human Genome Epidemiology Network (HuGENet): making sense of 10,000,000 postulated genetic risk factors John P.A. Ioannidis Department of Hygiene and Epidemiology University of Ioannina School of Medicine, Ioannina, Greece and Tufts University School of Medicine, Boston, USA

Human Genome Epidemiology Network (HuGENet) Global collaboration of individuals and organizations to assess population impact of genomics and how it can be used to improve health and prevent disease

Global collaboration of individuals and organizations to assess population impact of genomics and how it can be used to improve health and prevent disease

Genotype prevalence Gene - disease association Gene - gene interactions Gene - environment interactions Assessment of Genetic tests “ Systematic application of epidemiologic methods and approaches to assess the impact of human genetic variation on health and disease” Khoury, Little and Burke, HuGE 2004 HuGE problem: 25,000 genes, their combinations and interactions with risk factors

Genotype prevalence

Gene - disease association

Gene - gene interactions

Gene - environment interactions

Assessment of Genetic tests

From Genetics to Genomics Genetic Disorders Mendelian Disorders Disease burden: 5% Mutations/One Gene High Disease Risk Environment +/- “ Genetic Services” Genetic Information All Diseases Disease Burden: 95% Variants/MultiGenes Low Disease Risk Environment +++ General Practice

Genetic Disorders

Mendelian Disorders

Disease burden: 5%

Mutations/One Gene

High Disease Risk

Environment +/-

“ Genetic Services”

Genetic Information

All Diseases

Disease Burden: 95%

Variants/MultiGenes

Low Disease Risk

Environment +++

General Practice

Counting fish in the sea of gene-disease associations Ioannidis, Trends Mol Med 2003 Candidate analyses 1 quadrillion Investigators >10 Genetic contrasts >10 Subgroups >10 Outcomes >10 Diseases >1000 Gene variants >10000000 Parameter Multiplier

Major postulated problems of human genome epidemiology Small sample sizes Small effect sizes Large number of biological factors Main effects vs. interactions of genes Interaction with environmental exposures Potential variability across populations Old-epidemiology problems: confounding (population stratification), misclassification Questionable replication validity

Small sample sizes

Small effect sizes

Large number of biological factors

Main effects vs. interactions of genes

Interaction with environmental exposures

Potential variability across populations

Old-epidemiology problems: confounding (population stratification), misclassification

Questionable replication validity

Background issues Assay development Minimizing measurement error Minimizing design error Standardization and harmonization Validation Biological rationale and plausibility Clinical and public health use

Assay development

Minimizing measurement error

Minimizing design error

Standardization and harmonization

Validation

Biological rationale and plausibility

Clinical and public health use

Small sample size

Small genetic effects Ioannidis , Trikalinos, and Khoury , Am J Epidemiol 2006

Number of Published Human Genome Epidemiology (HuGE) Papers* 2001-2005 Year Prevalence Associations Interactions 2001 308 2141 436 2002 349 2799 569 2003 328 3021 600 2004 430 3772 664 2005 418 4569 911 * Data from HuGE Published Literature Database

Year Prevalence Associations Interactions

2001 308 2141 436

2002 349 2799 569

2003 328 3021 600

2004 430 3772 664

2005 418 4569 911

Epidemiologic Studies/Platforms Analyses and Publications Knowledge Integration Knowledge Dissemination Network of Networks/ P3G Guide Health Services & Inform Public Policy Guide Research Agenda The HuGENet “Road Map”

HuGENet “Network of Networks” Nat Genet Jan 2006

Examples of Network-based HuGE Study Platforms Disease Consortium Teams Subjects Parkinson GEO-PD 18 10,000 Osteoporosis GENOMOS 10 30,000 Preterm birth PREGENIA 10 20,000 Lymphoma INTERLYMPH 15 20,000 Lung cancer ILLCO 30 51,000 Head & Neck INHANCE 13 28,000 Melanoma GENOMEL 12 3,000 Pancreatic Ca PACGENE 10 5,000

Disease Consortium Teams Subjects

Parkinson GEO-PD 18 10,000

Osteoporosis GENOMOS 10 30,000

Preterm birth PREGENIA 10 20,000

Lymphoma INTERLYMPH 15 20,000

Lung cancer ILLCO 30 51,000

Head & Neck INHANCE 13 28,000

Melanoma GENOMEL 12 3,000

Pancreatic Ca PACGENE 10 5,000

Study Platforms: Methodologic Implications Assembling Teams Overall Project Design Harmonization vs standardization Outcome definitions and ascertainment Risk factor definitions and ascertainment Gene Selection and Measurement of genotypes Other biological markers Integrating and understanding the environmental variables (see. Davey-Smith et al. IJE 2006)

Assembling Teams

Overall Project Design

Harmonization vs standardization

Outcome definitions and ascertainment

Risk factor definitions and ascertainment

Gene Selection and Measurement of genotypes

Other biological markers

Integrating and understanding the environmental variables (see. Davey-Smith et al. IJE 2006)

Non-replicated diminishing effects Ioannidis et al , Nature Genetics 2001

Breast cancer: a “null” field for common genetic variants? Ioannidis , JNCI 2006 and Pharoah et al. JNCI 2006

The other side: don’t give up early

H: heterogeneity R/F: difference in first vs. subsequent D1-D3: small-study bias diagnostics RS/FS: significant findings (with/without first studies) Ioannidis et al, Lancet 2003

Succession of early extremes: the Proteus phenomenon Ioannidis and Trikalinos, J Clin Epidemiol 2005

Language bias and global science Pan et al. PLoS Med 2005

Measurement error: insight from a collaborative analysis Of 18 teams of investigators participating in the collaborative analysis of alpha-synuclein REP-I variation and Parkinson’s disease risk, we found that 7 had to be excluded from the main analyses because of laboratory error exceeding 10% and/or overt violation of HWE in the controls Two other teams who had published an inverse association apparently had miscoded the alleles in their databases. Maraganore et al , JAMA 2006

Of 18 teams of investigators participating in the collaborative analysis of alpha-synuclein REP-I variation and Parkinson’s disease risk, we found that 7 had to be excluded from the main analyses because of laboratory error exceeding 10% and/or overt violation of HWE in the controls

Two other teams who had published an inverse association apparently had miscoded the alleles in their databases.

Let us add the environment Hunter D. Nat Rev Genet 2005

Genomic Tests: a Public Health Issue Can potentially affect a lot of people (especially pharmacogenomics) Potential for enhancing and targeting prevention efforts Implementation and access Provider and public education Monitoring impact on population health

Can potentially affect a lot of people (especially pharmacogenomics)

Potential for enhancing and targeting prevention efforts

Implementation and access

Provider and public education

Monitoring impact on population health

“ Genomic profiling to promote a healthy lifestyle: not ready for prime time” Haga S et al. Nat Genet 2003

For each intended use A nalytic Validity C linical Validity C linical Utility Define test, disorder, and setting Knowledge Integration Across Disciplines: Evaluation of Genetic Tests http://www.cdc.gov/genomics/activities/fbr.htm

For each intended use

A nalytic Validity

C linical Validity

C linical Utility

Define test, disorder, and setting

Research on genetic risk factors Single studies: purely hypothesis-generating, important to register data, regardless of results Meta-analyses of group data: increasing certainty when several thousand subjects available Large-scale evidence from consortia or meta-analysis of individual-level data: evolving gold standard?

Single studies: purely hypothesis-generating, important to register data, regardless of results

Meta-analyses of group data: increasing certainty when several thousand subjects available

Large-scale evidence from consortia or meta-analysis of individual-level data: evolving gold standard?

Plenary 3 Co-chairs: Sally Green and Helen Whelton Pembroke