Transcript Overview of the September 2002 Operating Grants Competition

A Tale of (More Than ?) Two
Cohorts – from Canada
By Dr. John Frank, Scientific Director,
CIHR-Institute of Population & Public Health
Professor, Dept. of Public Health Sciences,
University of Toronto
Senior Scientist, Institute for Work and Health,
Toronto
November 16 -19, 2005
Why study gene-environment
interactions?
 Most disease burden is jointly determined
by interaction of individual genetic
endowments and complex sequence of
environmental factors
 These gene-environment interactions
require decades to fully manifest over the
life course
 Diseases and conditions of later life occur
in some and not others because of
intense interactions between particular
genetic constitutions and particular
sequence of social and physical
environments
Why study gene-environment
interactions? cont’d
 BUT…little is known about underlying causes
of these conditions and why they are now
increasing in frequency – for e.g. asthma
 Requires study of these sequential events in
large numbers of people over time, on whom
baseline genetic and repeated environmental
exposures are taken, to:
 understand the causal pathways; and,
 develop disease prevention strategies
Canadian “Contenders” in the
Great Cohort Competition

Canadian Longitudinal Study of Aging [CLSA] –
let by the CIHR – Institute of Aging

Canadian National (Multi-Generational) Birth
Cohort [CN/MBC] – led by the CIHR Institutes
of Human Development, Child and Youth
Health; Population and Public Health; and
Genetics
 Both of the above are also called the Canadian
Lifelong health Initiative [CLHI]

A “Canadian” Cancer / Chronic-Disease Cohort?
Canada’s scientific capacity for
a Birth Cohort Study

Disciplinary perspectives required in design and execution
of a cutting-edge birth cohort – for example: reproductive
and developmental biology, neurosciences, genetics,
epidemiology and biostatistics

Small size of scientific communities – help facilitate
formation of required trans-disciplinary teams

Determinants of health conceptual framework – provides
scientific context

Single payer health system

Capacity to link administrative datasets with interview and
examination data

Unique ‘environmental exposures’ in the Canadian context
Linking Adult and Child Cohorts

CLSA first “out of the gates”
 Protocol with 3 PIs, 200 Canadian collaborators
 International peer-review of protocol
 50,000 “tracking cohort” and 30,000 “intensive cohort”, age 40 +

CNBC focused on quantitative traits (height, weight,
adiposity, BP, lipids, [IgE]/FEV1, attentiveness) for
sample size reasons (n = 30 to 50,000) given costs per
subject

Linking CLSA and CNBC into a multi-generational cohort
suggested at CNBC meeting

Excited Canadian geneticists and international experts
International Workshop
on a Birth Cohort,
Toronto, Feb. 2003
A Multigenerational
Birth Cohort
Study
Linking Adult and Child Cohorts

Advantages
 Easier to “market”
 Children can be link to
recruiting grandparents
 Exposures and DNA over 3
generations
 Ascertain intermediate
phenotypes on parents
 Can supplement with
sample of older adults
without grandchildren

Disadvantages
 Excludes elderly parents
and therefore children of
recent adult immigrants
 Excludes children of
grandparents who died
prematurely if CLSA is
sampling from CNBC
 Timing / financing
challenges
Cancer/Chronic Disease Cohort

Cancer community convinced of necessity
 Emphasis on modifiable environmental and life-style risk factors
 Difficult/impossible to measure retrospectively (case-control)

EPIC, Biobank very expensive, payoff questionable?

Other Institutes/communities “not satisfied with”
chronic disease aspects of CLSA
 Cardiovascular disease
 Neurological disease
 Arthritis and other musculoskeletal disease

Canada could participate in international cancer
cohort study (John Potter’s “Last Cohort”; n = 106)
Studying Genetic and Environmental Contributions to
Disease Causation: An Uneven Playing Field
Measurement
Attribute
Genetic Exposure Measures
Environmental Exposure
Measures
Time-varying?
No – one sample per lifetime is enough
(unless gene expression arrays are
used)
Yes – new samples needed
whenever exposure changes
Data Collection Costs
Cheap (on a sample)
Expensive (real-time assays)
Sample Storage
(for later analysis)
Easy (buccal swab, buffy coat)
Difficult (e.g. air/water/diet
samples)
Data Analysis Costs
Getting cheaper by the day
Getting Costlier (as awareness of
chemical/physical/biological
complexity increases)
Overall Ease & Cost of
Accurate Ascertainment
Easy / Cheap
Difficult / Costly
Comparison of “Huge, Data-Thin” Cohorts (e.g. U.K. BioBank)
And “Small, Data-Thick” Cohorts (e.g. Southampton)
Cohort
Attribute
Huge – Thin
Small – Thick
Cost Per Subject
due to:
Low
(e.g. < $500. / data-wave)
High
(if > $1,000. / data-wave)
Sample Size
due to choice of:
500,000+
< 30,000
Exposures
Cheap-to-collect/store measures – e.g.
genetic
Expensive, balanced mix of
environmental and genetic
measures
Outcomes
Cheap-to-collect administrative data –
e.g. hospitalizations for
diagnoses/deaths
(dichotomous)   SS.
Expensive, directly measured bichemical physiologic, imaging,
functional outcomes
(often continuous)   SS.
Leading “ExposureMeasure Bias”
Large environmental exposure error
>> genetic factor errors
“Better balanced errors” for
environmental versus genetic
factors
Leading to:
Biased main effects and
interaction results
Less biased results
How Many Cohorts?

Can Canada fund more than one large national
cohort study?

Sample size dictated by rarest outcomes
 100,000-300,000 too small even for common cancers in
adults
 30,000-50,000 sufficient for healthy aging and most chronic
diseases in adults, and for detecting gene-environment
interactions influencing quantitative traits in both adults
and children

Exposures often outcome-specific for children
 Pre- and peri-conceptional exposures and pregnancy outcome
 Indoor air exposures and asthma
 Maternal-child interaction and ADHD or conduct disorder (CD)
Canadian Pregnancy/Birth Cohort:
Remaining Questions
1)
Is a cohort study of primarily quantitative traits of major scientific
value, or must dichotomous diseases be the primary outcomes?
2)
Can a single, omnibus cohort study address the exposures and
outcomes most important for Canadian mothers and children?
3)
Are we better off with several very small purpose-built cohort
studies of specific outcomes and exposures important to child
health (like IHDCYH’s recently-launched RFA on asthma)?
4)
Should we become the “Canadarm” of the U.S.’s NCS? At what
level of investment / sample size (10% = 10,000)?
5)
What is the precise scientific value of a multi-generational study?
6)
Should the CLSA wait until the CNBC “catches up”? What are the
resultant risks and costs?
Funding Issues

Can Canadian contribution to international cancer cohort
be funded from existing sources? (n = 50,000?
100,000?)

CLSA requires substantial resources outside of IA’s (and
CIHR’s) current budget – what to do?

Purpose-built pregnancy/birth cohort studies for powerful
gene-environmental interaction modeling (n = 30,000 +)
cannot be modeled on IHDCYH’s asthma RFA
 Partners like Allergen, NRC, and CHMC not readily available
for pre-conception exposures and ADHD/CD
 Cannot realistically be funded from any individual Institute’s
$8 m annual budget (or CIHR’s current budget of $700 m,
annually) given other pressures on them.
Next Steps

Meeting in Toronto on December 8-9
 Distinguished international advisory panel
 Canadian experts in reproductive/child health,
genetics, aging, and cancer
 Potential federal partners (StatsCan, Public Health
Agency of Canada, Social Science and Health Research
Council, Health Canada)
 Discuss Canada’s scientific “cohort niche”
 Panel to advise CIHR on best strategy(ies) to pursue

Decision by CIHR President, VPs, and SDs

Decision by CIHR Governing Council

Joint (with partners) “ask” for additional funding to ?
Federal Cabinet – new “Big Science Review” Process