The COHORTS collaboration - CHICOS project · The COHORTS collaboration Birth cohort research in low-and middle-income countries Linda Richter on behalf of COHORTS ... Challenges

The COHORTS collaborationBirth cohort research in low-

and middle-income countriesLinda Richter

on behalf of COHORTS

CHICOS, Barcelona 11-12th April 2011

FIRSTLY, COHORTS is an informal affiliation between a number of researchers with

shared interests in:• Nutrition and public health in LMI countries• Birth cohort studies• Developmental origins of health and wellbeing

and• A mix of complementary skills

now also• Strong friendships across sites

What is COHORTS?

SECONDLY, COHORTS is a formal, funded network between 5 birth cohort studies

• With a project leader• Paid data coordinators in each site • Regular meetings• Workplans and schedules• Joint publications• Student and staff exchanges

How does COHORTS work?

How did COHORTS begin?

Lancet series

Maternal and Child UndernutritionLaunched in London, 16 January 2008

Maternal and Child Undernutrition Study Group

Led by Cesar Victora – Pelotasassisted by Pedro Hallal

• Rey Martorell – Guatemala / Emory• Linda Adair – Philippines / North Carolina• Harshi Sachdev – New Delhi• Linda Richter – Soweto / Jhb

Invited collaborators

• Exposure variables - growth o during pregnancy (maternal height and weight)o before pregnancy (weight gain, micronutrient status and

diet)o at birth (weight, length, ponderal index, intrauterine growth

restriction), and o 2 years of age (stunting, wasting, underweight)

• Outcome variableso 14 adult outcomes: height; achieved schooling & educational

performance; income and assets; birthweight in the offspring; BMI, body composition, and obesity; blood lipids; insulin resistance & type 2 diabetes; blood pressure; cardiovascular disease; lung function; immune function; cancers; bone mass, fracture risk, and osteoporosis; and mental illness

Systematic review

• Exposure variables o Maternal height o Birthweight, IGR (BW for GA <10th percentile, sex specific)o 2 years – HFA /WFA /BMI z-scores; stunting, underweight

wasting (z-score cut-off -2)

• Adult outcome variableso Height, achieved schoolingo Offspring birthweighto BMI, blood glucose, systolic blood pressure

• Potential confounding variables o Age at outcome o Parental schoolingo Early childhood socioeconomic status (household assets).

Measured variables

First COHORTS dataset !

• Combined data – meta-analytic procedures• Striking similarity – sites and gender• Strong evidence - adequate nutrition in utero

& first 2 years of life human capital. Undernourished children are more likely to:

– Be short adults – Have lower educational achievement, earnings – Give birth to smaller infants, also in 3rd G

• Chronic disease results not so straightforward – but rapid growth under 2ynot related

Data analysis & results

Results: Eg height

1 z-score at 2 years = 3.2cm taller in adulthood

COHORTS

COHORTS activities

• 2006-7 (Lancet nutrition series, 2008)– Birth weight and size at 2 yr in relation to young adult

health and human capital• 2007-9 (Wellcome Trust Grant – C Victora)

– Risks associated with initial poor growth and rapid weight gain at different ages? Infant feeding

patterns?• 2010-12 (Wellcome Trust Grant – L Richter)

– Long-term effects of early physical and social environments

• 2010-2011 (Gates Foundation Grant – A Stein)– Evidence for interventions in first 1 000 days

LMIC countries? • Understudied relative to high income countries• Currently experiencing dramatic increases in

chronic disease• Today’s adults were born when early life

nutritional deficits were more common (high rates of LBW, IUGR, and early childhood stunting)

• Confounders are differently associated with exposures and outcomes compared to developed countries

High Income Countries

Low Income Countries

Obesity poverty wealth

Breastfeeding

high education low education

Early weight gain?• In high income settings, rapid infant and young child

weight gain elevated risk of obesity and type 2 diabetes later in life

• In low and middle income countries, rapid weight gain following a period of poor nutrition may be essential for survival and prevention of impaired cognitive development

• Information is needed to weigh the short and long term risks in different settings

Networking advantages

• Scientific rationaleo Increased sample sizeo Variation in frequencies of exposures and outcomeso Varying nature of some exposureso Different confounding patternso Less publication biaso Cross-fertilization, inter-disciplinarityo Capture life trajectories, demographic/health

transitionso Greater visibility, impact on policy (?)

• Capacity enhancemento Methodological and statistical capacityo Student exchanges

Size - commonalities

• Population-based samples in low and middle-income countries

• Recruited during pregnancy or at birth• Long-term follow up (>20 yr)• Repeated visits in early postnatal life

– Social, environmental, demographic and related variables

– Anthropometric measurements– Feeding patterns, morbidity and mortality

• Assessments in young adulthood– Human capital (schooling, income)– Health related behaviors (smoking, physical activity)– Precursors for chronic diseases (body composition,

blood pressure, glucose)

Eg Variation in exposures

Prevalence of low birth weight and stunting at age 2 years in the 5 cohorts

Eg Variation in confounders

• Different ages at enrolment & current• Different ages at follow-up data collection• Availability of variables• Different definitions of variables• Different lab/collection methods• Lack of data on confounding variables• Different start times, transitions (1969

Guatemala, India 1982, 1983 Pelotas, Cebu 1990 Soweto)

Challenges of joint analyses

5 cohortsCOHORT ENROL FOLLOW-

UPN

Orig, last analAGE AT FOLL-UP

New Delhi Pregnancy(1969-1972)

1sty (3m’ly)1-20y (6mly)29-32, 37-40

8 1811 149 29.1

Guatemala Birth(1969-1977)

11-19, 25-30, 25-35, 28-38

2 3921 571 29.6

Pelotas Birth(1982)

2,4, 23 yrsSubs 1,15, 18,19, 26

5 9144 297 22.7

Cebu Pregnancy(1983-1984)

0-2y (2mly)8,11,15,19,

22y

3 0802 032 21.3

Soweto/ Jhb Pregnancy(1990)

0, 6mo1,2,3,4,5,7,11, 13-19 (2*yly)

3 2731 869 19.5

• Drs Shanti Ghosh & Santosh Bhargava (1969)

• 12km2 area of South Delhi• 20 755 married women of reproductive age,

followed bi-monthly to record menstrual dates• 9 169 pregnancies 8 181 live births• Recorded wt, ht at birth, 3 and then 6mthly till 20y• Follow ups at ±30 and ± 40yrs

New Delhi

• Drs Rey Martorell, Aryeh Stein and others (1969)

• Large-scale randomised intervention (“Atole”) trial• Screened 300 villages, selected 4 comparable• 2 392 eligible children – <7y or born (1969-1977)• Follow-up to ± 40yrs, +200 publications

Guatemala (INCAP)

2008

• Drs Fernando Barros and Cesar Victora (1982)

• 3 overlapping birth cohorts – 1982, 1993 & 2004• 1982 – 3 maternity hospitals, 5 914 live births

recorded• Early follow-ups incomplete – inadequate funding• Now well-planned and funded (Wellcome a.o.)

Pelotas

• Drs Linda Adair, Barry Popkin & others (1983)

• Cluster sampling – 17 urban & 16 rural barangays• 2 800 households – 3 080 eligible singleton births• Tracking 2mthly to 2yrs, 7 follow ups to 28yrs• ±130 papers

Cebu, Philippines

• Drs Linda Richter, Shane Norris & others (1990)

• Mandela’s Children – 7 wks following NM’s release• Monitored antenatal attendance in all city clinics• Enrolled from 6m pregnancy, 3 273 births• Since 13yrs, followed up twice a year, incl HIV

Soweto-Johannesburg

COHORTS

2011

Papers 2008-2010

• COHORTS (Richter et al 2011, Int J Epidemiol)

• Human capital– Adult height (Stein et al 2010, Am J Hum Biol)– Achieved schooling (Martorell et al 2010, J Nutr)

• Precursors of complex chronic diseases– Blood pressure (Adair et al 2009, Am J Clin Nutr)– Body composition (Kuzawa et al, in progress) – Glucose (Norris et al, submitted Diab Care)

• Long-term consequences of infant feeding – (Fall et al 2010, Int J Epidemiol)

• Methodological issues in studies of early growth

– (Osmond et al, submitted J DoHaD)

Growth - conclusions• Early childhood wt gains are less detrimental than

late childhood/adolescent wt gains, and can even be beneficial for some outcomes…likely because early weight gain is related to higher lean mass

• Later wt gains are consistently associated with adverse outcomes… likely because they are associated with greater gains in fat mass

• Any detrimental effects of early wt gain on adult chronic disease must be interpreted in light of the beneficial effects of wt gain for survival, short-term morbidity and for long-term gains in human capital in LMICs

Schooling

•Fertility, child health and nutrition

•Gender equity and empowerment

•Long term health benefits to individuals

•Earnings

Latin America & Caribbean

12.0%

Sub-Saharan Africa

11.7%

Asia 9.9%

World average: 10.0%

The average rate of return to income of another year of

schooling

Psacharopoulos G, Patrinos HA. Returns to investment in education: A further update. Educ Econ 2004; 12(2): 111–134.

Importance of schooling

Schooling analysis

• Used conditional weight/height gain variables (0-2 and 2-4 y), uncorrelated with each other or birthweight

• Tested for interactions with sex, found none. Pooled results for males and females

• No significant interactions by site for highest grade attained and ever failed a grade. Pooled results by site

• Significant site interactions for age at school entry - analyses were stratified by site

Increased years schoolingYrs increased schooling associated with a SD difference in birth

wt and conditional wt gain (0-2y and 2-4y)

Age at school entry

• Results heterogeneous

• No significant associations in Guatemala

• In general, relationships in Brazil, the Philippines and South Africa indicate

o Association between larger birthweights and faster weight gains with younger ages at enrollment

Schooling - summary

For each 1 SD shift in conditional weight at age 2 (ie faster gain from 0-2 y):– Attained schooling increases by 0.43 yr– Likelihood of ever failing a grade is reduced by

12%– Benefits are greater for smaller children

Models adjusted for household SES in childhood, maternal schooling, age, sex, site

Martorell et al. 2010. Weight gain in the first two years of life is an important predictor of schooling outcomes in pooled analyses from five birth cohorts from low- and middle-income countries. Journal of Nutrition , 140, 348-354.

Papers in progress (Gates)Strengthen evidence for interventions in window of

vulnerability/opportunity – first 1000 days• Impact of preterm birth on growth, human health &

capital• Characterize healthy postnatal growth

o rapid wt vs ht gain after stunting at 2yo early growth and timing of pubertal development in girls

• Intergenerational determinants of growtho poverty & low parental schooling on growth < and > 2yo intergenerational constraints on birth size and infant growth

• Relation between breastfeeding duration and achieved schooling

COHORTS 2010-2012

Long-lasting effects of early physical and social environments• Physical and social environments in early

childhood and growth in the first 2yrs of life?• Early environmental influences on adult health

and human capacity?• How physical and social environments in early

childhood moderate the relationships between growth in the first 2yrs of life and adult health and capacity?

Bronfenbrenner’s bio-ecological model of evironmental influences on children (1979)

CHILDsex, birth order

MOTHERAge at 1st birth, height, education, “authority”

HOUSEHOLDDensity, dependency, marital status, family type, father presence

ECONOMICAssets, income, wealth, employment

INFRASTRUCTURE & SERVICESAccess & use of services, water, sanitation

Early environments

• Systematic literature reviews of each component - child, mother, household, infrastructure, economic – on:

o Early growtho Adult health and capacity

• Construction of database for analysiso Harmonise variables, age rangeso Standardize variableso Create comprehensive codebook

• Overview paper + individual papers

Practical arrangements

• COHORTS Project leader• Coordinating group – 1 investigator per cohort

o Decide on priority analyses and publicationso Data availabililtyo Publication guidelines

• 1 lead site/investigator for analysis and authorship of each paper

• All papers authored by o writing group (up to 3 investigators, from different sites) o + 5 PIs (or their representatives)o + COHORTS group

Practical arrangements #2

• Annual meetings of PIs, 1-2 senior investigators, 1-2 junior investigator per cohort

• 1-2 small-group working meetings per year• Every 2 years coordinated with DoHaD• Core dataset available to all• Additional variables available upon request• Parallel network of junior investigators /

fellowships (exchange, training)

Main achievements

• First-rate group of researchers• Remarkable openness and willingness to

share data• Rotating authorship and coordination

structure• Production of high-visibility research• Learning from each other and building

capacity of young researchers

Main challenges

• Technical issues of data analysiso Meta-analyses or pooled analyses? Raw scores

or z-scores? etc• Dealing with heterogeneity, confounding• Handling losses to follow-up, missing data• Finding solutions to differences in exposure

and outcome variables, different measurements, different time periods

• Developing a core data set

Next steps

• Maintain pace of shared publications• Ensure long-term funding• Build stable high-level capacity• Coordinate new data collection• Set up new cohorts with appropriate early

life measurements? – eg Soweto 3G enrolment

Conclusions

• COHORTS represents a unique resource in LMIC countries

• COHORTS is a model for successful data sharing and pooling, collaboration

o Active engagement by senior and junior investigators

o Exploring opportunities for common data collection protocols

o Linkages with other cohorts (eg ALSPAC)

Conditional size

• The high correlation of repeated measures of size (wt, ht) during childhood a challenge for analysis

• We modeled “conditional weight ” (CW)o CW = residual from site- and sex-stratified

regressions of wt (kg) at a given age on all prior wts

o CW at any age is uncorrelated with CW at prior ages• CW at any age represents deviation of child’s

wt from its expected value • When included together in a multiple regression,

CWs can be interpreted as relative change in wt during the prior interval

Globalization

• Transitions– Nutrition– Morbidity– Physical activity– Leisure patterns– Behaviour

• But also improvements in– Health care– Education– Water/sanitation– Etc