Upload
others
View
5
Download
0
Embed Size (px)
Citation preview
Lifecourse Epidemiology, Longitudinal Studies
and
Knowledge Transla:on
John Lynch Professor of Public Health and Epidemiology
University of Adelaide
NHMRC Australia Fellow
hHp://www.abc.net.au/tv/life/about_the_series/life_at_5.htm
Life at 5
Lifecourse Epidemiology
Birth Old Age Middle Age Childhood Adolescence
Perinatal Epidemiology
Birth Outcomes
Adult Risk Factor Epidemiology
Adult Chronic Diseases
Lifecourse Epidemiology
The childhood shows the man,
As the morning shows the day.
Milton “Paradise Lost” (1667)
lines 220–21
Not a new idea
Died ~ 1920
Died ~ 1900
Widdowson and McCance (1963) showed that rat pups with limited nutri:on
during the 3 weeks of lacta:on put on weight more slowly over their en:re
life:mes, even though they had unlimited food post-‐weaning.
In contrast, in another groups of rats, the same dura:on of limited nutri:on at
9-‐12 weeks of age had only short term effects on their weight.
Widdowson and McCance. Proc R Soc Lond B Biol Sci (1963)
Evidence for “cri:cal periods” of “biological programming”
Rediscovering Lifecourse
Dubos. Pediatrics (1966) Dubos, et al. Pediatrics (1966)
Atherosclerosis, most frequently manifested as coronary heart disease, is the major
health concern in the United States. This disease of adulthood has its origins in
infancy and childhood.
Kannel WB, Dawber TR. Atherosclerosis as a pediatric problem.
Journal of Pediatrics 1972: 80; 544-‐554
September 2010
The Barker Hypothesis
The “Developmental Origins” Hypothesis
hHp://www.barker.org/
Men Women Birthweight <5.5 1.0 1.0 6 -‐ 0.81 0.87 7 -‐ 0.80 0.81 8 -‐ 0.74 0.71 9 -‐ 0.55 0.52 10 + 0.65 0.59
Osmond, et al BMJ (1993)
Rate Ra:os for CHD in Rela:on to Birthweight
Weight at 1 year <18 1.0 1.0 19 -‐ 0.79 0.59 21 -‐ 0.81 0.75 23 -‐ 0.62 0.52 25 -‐ 0.62 0.84 27 + 0.40 0.84
Barker et al. NEJM (2005)
Trajectories of Growth Among Children Who Have Coronary Events as Adults
3 “cri:cal” phases:
Lower birthweight Sub-‐op:mal growth to age 2 Catch up growth to adolescence
Girls Boys
Di Kuh and Yoav Ben Shlomo (2004)
Developmental Plas:city and Human Health
Bateson, Barker, et al. Nature (2004)
• In-‐utero responsiveness to their mothers’
condi:on before birth may generally
prepare individuals so that they are best
suited to the environment forecast by cues
available in early life.
• Rapid improvements in nutri:on and
other environmental condi:ons may have
damaging effects on the health of those
people whose parents and grandparents
lived in impoverished condi:ons.
• A fuller understanding of paHerns of human plas:city in response to early nutri:on and
other environmental factors will have implica:ons for the administra:on of public health.
2006
JE Harding
The nutri:onal basis of the fetal origins of adult disease
Int. J. Epidemiol. 2001 30(1): 15-‐23
2001
Gluckman and Hanson
Sundström, et al. BMJ (2011)
Birth weight and Cogni:ve Func:on
Richards, et al. BMJ (2001)
BW = Cogni:ve Func:on @ ages 8, 11, 15, 26
Survival at ages 12-‐76 years by IQ score at age 11 in UK
Whalley & Deary, BMJ 2001
Female Male
Early Life Conditions
Health in Adulthood
Adult Socioeconomic
Position
Adult Exposures
Biological Processes
“Health Capital”
Economic Resources D
Educational Experiences
C
Exposures in-‐utero
and childhood
A Health Behaviours
B
Kuh, Power, Blane, Bartley. (2005)
Child and Adolescent Health
E
• 1 kg difference in birthweight 4mm Hg difference in systolic BP
• Difference in BW between smoking/non-‐smoking mothers ~ 200 gm
• Protein supplementa:on among under-‐nourished women 50 gm increase BW
• More realis:c change is 100 gm BW improvement 0.4 mmHg
“Indeed, for most of the epidemiological associa:ons described to date, the
extent to which interven:ons that are intended to modify risk can improve
long-‐term health is not yet clear.”
Gillman M. Developmental Origins of Health and Disease. NEJM (2005)
Death Rates from CHD and Stroke, US, 1950-‐2002
Stroke
CHD
Rate Per 100,000
70%
70%
US Birthweight Distribu:on 1950-‐2000
Longitudinal Studies
1937-‐1939
“In 1926-‐1927 an experiment with 1500 children done by the
RoweH Ins:tute showed that the addi:on of milk made a definite
improvement in the rate of growth and health in ordinary children
… milk made good the dietary deficiencies in their home diet which
was worst in the poorest homes.”
Lord Boyd Orr
Dr John Boyd Orr
Nature March 2011
The first few decades of the twen:eth century found Britain acutely concerned about its
falling birth rate and stagnant infant mortality – “how are we going to maintain Britain
and its empire?”
A Popula:on Inves:ga:on CommiHee recommended a maternity survey to explore
whether the social and economic costs of childbearing were discouraging prospec:ve
parents. James Douglas was appointed to head it.
Mike Wadsworth, a social epidemiologist joined Douglas’s team in 1968 – the study was
just gewng going. “I thought the changing paHern of health of these people would be
interes:ng over life,” he says.
Di Kuh — who had trained in economics — wanted to build up the biomedical data
that Wadsworth had been collec:ng. Un:l that :me, all the examina:ons had been
performed at the study members’ homes, but by this stage the nurses were
staggering under all the equipment.
To really understand the par:cipants’ physiology and biology, Kuh argued, the study
needed to get them to a clinic. “People appreciate a free bone scan,” she said. By
2008 she had convinced the MRC to pay for every willing cohort member to visit
one of a number of clinics around the country.
The UK Birth Cohorts
• 1946
• 1958 • 1970
• Millenium Cohort
• 2012 Cohort
Perinatal death and health care
Infant feeding
Educa:on and schooling transi:ons
Secular socioeconomic change – employment, family structure
Equity and geographic varia:on
Trajectories linking early life and later health
1959 – PlaH Comm. on Welfare of Children in Hospital
1966 – Finer Comm. On Single Parent Families
1967 – Plowden Comm. on Primary Educa:on
1978 – Warnock Comm. On Special Needs
1981 – Black Report
1998 – Acheson Report on Social Inequality
1999 – Moser Comm. On Basic Skills
Original Hypotheses (1948)
1. CVD increases with age. It occurs earlier and more frequently in males.
2. Persons with hypertension develop CVD at a greater rate than those who are not hypertensive.
3. Elevated blood cholesterol level is associated with an increased risk of CVD.
4. Tobacco smoking is associated with an increased occurrence of CVD.
5. Habitual use of alcohol is associated with increased incidence of CVD.
6. Increased physical ac:vity is associated with a decrease in the development of CVD.
7. An increase in thyroid func:on is associated with a decrease in the development of CVD.
8. A high blood hemoglobin are associated with an increased rate of the development of CVD.
9. An increase in body weight predisposes to CVD.
10. There is an increased rate of the development of CVD in people with diabetes mellitus.
11. There is a higher incidence of CVD in people with gout.
Dr. Thomas Dawber
American J Public Health. 1951;14:279-‐286
The study is focused on arteriosclero:c and hypertensive heart disease
because these are the most important of the cardiovascular diseases and
the least is known about their epidemiology.
As a working hypothesis it is assumed that these diseases do not each have
a single cause (as is the case in most infec:ous diseases), but that they are
the result of mul:ple causes which work slowly within the individual.
Terman Lifecycle Study (1922-‐1986) began by comparing a group of children with high IQ
(n=1470) with groups of children typical of the general popula:on, to discover similari:es
and differences. Research was con:nued over 60 years.
Wisconsin Longitudinal Study (1957 -‐ ) is a long-‐term study of a random sample of 10,317
men and women who graduated from Wisconsin high schools in 1957. The WLS provides an
opportunity to study the life course, intergenera:onal transfers and rela:onships, family
func:oning, physical and mental health and well-‐being, and morbidity and mortality from
late adolescence through 2008.
Panel Study of Income Dynamics (1968 -‐ ) examines income, behaviour and demographic
change of about 18,000 individuals in 4,800 families. Now includes inter-‐genera:onal
component.
Longitudinal Studies with Roots in Social Science
Dunedin (1973 -‐ )
The Dunedin Mul:disciplinary Health and Development Study involves 1037 babies born
in Dunedin between April 1972 and March 1973 at the Queen Mary Maternity Hospital.
Of those 1037 babies, 1014 of the original cohort are s:ll alive today.
The babies were first followed up at the age of 3, and then at 5, 7, 9, 11, 13, 15, 18, 21, 26
and 32. Future assessments are scheduled for age 38 (2010-‐2012), 44 and on into the
future as study members have their own families, age, and re:re.
Christchurch (1977 -‐ )
The Christchurch Health and Development Study began in 1977 and followed the health,
educa:on and life progress of a group of 1,265 children born in Christchurch. The data
gathered over the course of the study now comprises some 50 million characters of
informa:on describing the life history of this cohort.
Both these studies have achieved remarkable levels of follow-‐up of parLcipants
New Zealand Longitudinal Studies
Na:onal Longitudinal Survey of Children and Youth (1994-‐2009)
The Na:onal Longitudinal Survey of Children and Youth (NLSCY) was a long-‐term study
of Canadian children that followed their development and well-‐being from birth to
early adulthood.
The NLSCY began in 1994 with children aged 0-‐11 (n= ~16,900) and was jointly
conducted by Sta:s:cs Canada and Human Resources Development Canada (HRDC).
Cycle 8 was completed in 2009. AHri:on rates are ~ 25-‐30%.
It is no longer ac:ve.
Canada
Growing Up in Australia -‐ Longitudinal Study of Australian Children (2003 -‐ )
The study addresses a range of research ques:ons about paren:ng, family rela:onships,
childhood educa:on, non-‐parental child care and health. The study will further
understanding of child and adolescent development, inform social policy debate, and
will be used to iden:fy opportuni:es for early interven:on and preven:on strategies in
policy areas concerning children and families.
LSAC involves data collected from two cohorts every two years. The first cohort of 5000
children was aged 0-‐1 years in 2003-‐2004, and the second cohort of 5000 children was
aged 4-‐5 years in 2003-‐2004. Study informants include the child (when of an appropriate
age) and parents (both resident and non-‐resident), carers and teachers.
Ini:al response rate 50-‐55%
Australia
Na:onal Children’s Study (USA)
The Na:onal Children’s Study is a mul:-‐centre cohort that will aHempt to recruit in pre-‐
pregnancy and result in 100,000 live births. Planning began in 1999. It will examine the
effects of the environment -‐ broadly defined -‐ air, water, diet, sound, family dynamics,
community and cultural influences, and gene:cs on the growth, development, and health
of children across the United States, following them from before birth un:l age 21 years.
Hundreds of scien:sts and representa:ves from community groups and professional
organiza:ons have contributed to the iden:fica:on of key children’s environmental
health ques:ons for the Study.
The Study Design Working Group cons:tuted in 1999 proposed the development of core
hypotheses encompassing exposures and child health outcomes of great public health
significance requiring long-‐term follow-‐up and which cannot be reasonably studied with
fewer children or a different study design.
★ ★
★
★
★ ★
★
★ ★
★
★
★ ★
★
★
★
★
★
★ ★
★
★
★
★
★
★
★
★
★
★
★
★
★
★
★ ★
★
39 Birth Cohorts
Growing Up in New Zealand (2009 -‐ )
N = 6822 mothers and 4,404 partners
“ … Growing Up in New Zealand is designed to give us a more complete picture of the
pathways that lead to differen:al outcomes, and to provide much beHer evidence for the
development of strategies to reduce inequali:es and to improve outcomes for all children.”
1. What determines developmental trajectories across mul:ple levels of influence (poli:cal, social, cultural, intergenera:onal, familial, and individual) through the life course in the major domains of interest?
2. How are New Zealand children faring developmentally across mul:ple domains at discrete points in the lifecourse?
3. How are these developmental trajectories and outcomes associated with ethnicity across the lifecourse?
4. What factors and trajectories, par:cularly across mul:ple levels of influence, confer resilience and op:mise development for New Zealand children?
• 1 in 3 of our children is born to at least one parent who did not grow up in New Zealand
• for 1 in 10 children the rela:onship between their parents will change significantly between the beginning of the pregnancy and their birth
• increasingly our children are being born into varied family structures including mothers
living with extended families (nearly 24% overall and more than half of all Pacific families),
with non-‐kin (3% overall but greater than 10% for Asian mothers) or without other adults
(3% overall but 7% for Māori mothers).
• our families are highly mobile with over half of all families moving more than twice in the
last five years
• many parents in the most deprived areas were not aware of either Paid Parental Leave or
Working for Families tax credits.
Shulruf et al (2007)
NZ
Oz
Longitudinal Studies – Poten:al and Challenges
Lancet (2008)
“The record of achievement toward the goals of these public policy ini:a:ves
requires con:nuous replenishing of the evidence base from a new birth cohort as
the circumstances of birth and infancy change.”
Longview Report for the ESRC. The scien:fic case for a new birth cohort (2007)
Marmot Review (2009)
UK Health Inequali:es
Based on Feinstein (2006)
Age 5 Age 10
Challenges: Policy and Prac:ce Relevant
• Close rela:onships with policy/prac:ce partners
• For researchers -‐ framing policy relevant ques:ons
• For policy makers -‐ framing scien:fically “answerable” ques:ons
• Efficacy (RCTs) and effec:veness (pragma:c) trials – embedded in cohorts
• Asking ques:ons that can help inform prac:ce
Young maternal age and poor child development outcomes by age 5: findings on
predicJve validity from a longitudinal birth cohort
Catherine ChiHleborough, Debbie Lawlor, John Lynch. Pediatrics (in press)
0
10
20
30
40
50
60
70
80
Age <20 years No partner or not cohabitating
Financial difficulties
Depression Smoking Education less than O level
At least 1 of the six predictors
At least 2 of the six predictors
ALSPAC Dev Scale @ 18 mths
SDQ @ 47 mths
Std Educa:onal Assessment @ 60 mths school entry
% Cases in BoH
om 10t
h
51% 23%
ALSPAC (2011)
AHri:on and Missing Data
• Sample replenishment
• Complete case analysis?
• Mul:ple imputa:on
Individuals Q1 T1 Q2 T1 Q3 T1 Q4 T1 Q5 T1 Q1 T2 Q2 T2 Q3 T2 Q4 T2 Q5 T2
DNR
R T1 DNR Ts
DNR T1 R T2
Complete
C T1 M T2
M T1 M T2
1
2
3
4
5
Complex PaHerns of Missing Data
Aspira:onal Measurement and Measurement Error
• Desire to measure the dynamic complexity
• Measure some things poorly
• Measurement error may be the largest source of effect es:mate uncertainty
• InaHen:on to confounding – poorer causal inference
• Richly phenotyped sub-‐groups -‐ ALSPAC and asthma
• Biological data including gene:cs
• New measurement technologies – web, IT – iPhone apps, data linkage
Inter-‐disciplinarity
Lynch Int J Epidemiol (2006)
“It isn’t easy being inter-‐disciplinary”
2010
Produc:vity
1 per week
1 per day
Produc:vity – resourcing data collec:on + data management + use
Methodological Complexity and Causal Inference
• How do we fully exploit the rich data we collect over :me?
• Time dependent confounding
• Complex causal structures
• Innova:ve sta:s:cal methods
• Longitudinal studies are not just repeated cross-‐sec:ons
Intact (3590)
Single
Step
Single (306)
Step (84)
Intact (3200)
Intact (2744)
Single (130)
Step (38)
Step
Single
Step
Step
Single
Single (213)
Step
Step
Single
Single
Single (175) Step (38)
Step
Single (161)
Single
Intact
Intact (2912)
Intact
Intact
Age nine Age 11/12 Age five Birth
NLSCY (n = ~4000)
76%
Non-Intact (198)
Single
Step
Single (119)
Step (36)
Intact (43)
Intact
Single
Step
Step
Step
Single
Step
Step
Single
Single (74)
Step (18)
Step
Single
Single
Intact
Age 11/12 Age five Birth
“Would more sophis:cated sta:s:cal methods greatly clarify things? Somehow I
doubt it; however we should clearly try and evaluate whether they would …
Much of lifecourse epidemiology is focused on trajectories of the lives of individuals,
in which stochas:c processes that we may never be able to measure or understand
play an important role. Epidemiologists may be faced with a similar ‘gloomy prospect’
to the one iden:fied by Eric Turkheimer in the context of the inability of behavioural
gene:c studies to iden:fy important environmental factors shared between siblings
within families.
Lifecourse trajectory-‐influencing events may o�en be of chance or idiosyncra:c origin,
and thus not tractable by current methods.” Davey Smith. Int J Epidemiol (2007)
“… an exhausLve causal inves:ga:on of any concrete
phenomenon in its full reality is not only prac:cally
impossible -‐ it is simply nonsense …. The more “general”, i.e.,
the more abstract the laws, the less they can contribute to
the causal imputa:on of individual phenomena.”
Max Weber, 1904
Marmot Review (2009)
UK Health Inequali:es
Based on Feinstein (2006)
Age 5 Age 10
• Popula:on average processes
Knowledge Transla:on
My quesJon is: “Are we making an impact?”
Paré 1575
Trea:ng wounds with boiling oil
did more harm than good.
Ligatures were more effec:ve but
100 years before it was widely
accepted
Slow Transla:on?
• Scurvy Known since the early 1600s, Lind showed in 1747 how oranges and lemons eliminated scurvy among sailors but 1864 before Bri:sh Board of trade used citrus for all its sailors
• Smoking In 1950 Doll showed smoking and lung cancer linked but adver:zing not completely banned in UK (2002), US (2003), and Australia (1998)
• Asbestos In 1955 Doll demonstrated rela:onship between asbestos and lung cancer but asbestos manufacture not banned in Australia un:l 1987
• Childhood obesity The first data about increases in childhood obesity appeared in the 1960s, certainly by the 1980s but yet liHle systema:c policy response to quell the ‘obesogenic environment’
“lifecourse” or “life course” in the :tle
Publica:ons N=1523
Cita:ons N=18,217
H Index = 59
“returns on investment”
Research Produc:on Observa:onal selec:on bias confounding
measurement error
Experimental RCTs – not ethical , not prac:cal, not externally valid, efficacy focus
Pragma:c Trials – effec:veness focus, but who’s ques:on?
Natural experiments
Research Consump:on
Policy
Implementa:on in Prac:ce
Transla:on
The Gap
Co-‐crea:on
Informa:on Systems
Research Challenges
• Ask relevant research ques:ons
• Best evidence from high quality observa:onal studies
• Appropriate designs for valid / “good enough” evidence
• Funding – pragma:c trials unlikely to get perfect score on NHMRC criteria
• Reward structures – “of sausages and salami” (Konrad Jamrozik, MJA)
• Managed por�olios of ECD research vs Inves:gator driven
Policy and PracJce Challenges
• “Health in all policies” implica:ons for cross-‐sector integrated collabora:ons
• Integrated finance and management of inter-‐disciplinary teams
• Workforce training and workforce intransigence
• Openness to change
• A real desire to use informa:on to inform prac:ce
“Life is to be understood backwards, but it is lived forwards”
Søren Kierkegaard 1813-‐1855