challenges in harmonization and development of measures among the cohorts studies
TRANSCRIPT
Challenges in harmonization and development of measures among the
COHORTS studies
Linda AdairUniversity of North Carolina at Chapel Hill
Birth cohort publications
2
Source: www.worldmapper.orgWWW.WORLDMAPPER.ORG
WWW.GOPUBMED.COM (2009)
History of COHORTS
• Organized by Cesar Victora as a writing group for the Lancet Maternal and Child Undernutrition Series
• Motivated by lack of DOHaD studies based on birth cohorts in low and middle income countries
• Brought together the Principal Investigators of birth cohorts followed to late adolescence or adulthood
Theme for analyses:Early life influences on later health and
human capital formation
• Rationale for the selected outcomes– Cardiometabolic disease outcomes
• Diabetes and hypertension now contribute substantially to the burden of disease in low and middle income countries
– Body composition --- especial central body fat --- is strongly associated with diabetes and CVD risk and may mediate the relationship of early life nutritional status with CM disease risk
– Fasting blood glucose levels and impaired fasting glucose are important precursors to type II diabetes
– Systolic and diastolic blood pressure predict later hypertension and CVD
– Blood pressure, glucose, and body composition can be reliably measured in large population-based studies, and were available for a large portion of COHORTS participants
– Schooling and height are important measures of human capital: Potential conflicting effect of early weight gain on these outcomes relative to CM disease risk must be considered in LMIC
© 2009 - World Mapshttp://www.justmaps.org
Pelotas
INCAP: Guatemala
CLHNS Cebu
New Delhi
Birth to 20
Study CharacteristicsStudy Design Cohort
inception
Participants N with
BW and
adult BMI
Mean age
at follow-
up
Pelotas
Brazil
Prospective
cohort
1982 All infants born in the city’s maternity
hospitals (>99% of all births) during
1982. All social classes included.
4,148 22.7
INCAP
Nutrition
Trial Cohort
Guatemala
Community
trial
1969-77 Children <7 y in 1969 & those born
1969-77 enrolled, participants in an
intervention trial of a high-energy and
protein supplement in 4 rural villages.
544 29.6
New Delhi
Birth Cohort
Study
Prospective
cohort
1969-72 Infants born to married women from a
defined area of Delhi.
Primarily middle-class
1,583 29.1
Cebu
Longitudinal
Health &
Nutrition
Survey
Prospective
cohort
1983-84 One yr birth cohort from 33 randomly
selected communities of Metro Cebu;
75% urban. All social classes included.
2,001 21.3
Birth to 20
South
Africa
Prospective
cohort
1990 Infants from a delimited urban area
(Soweto, Johannesburg)
Predominantly poor, black sample.
1,567 15.6, age 18
data added
later
All sites exemplify discordance of birth and adult weight status
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1LBW Stunted (LAZ<-2) BMI>25 BMI>30
Males Females Males Females Males Females Males Females Males Females
Brazil Guatemala Delhi Cebu Soweto
Pro
po
rtio
n
Identifying directly comparable data
• Identify essential core data needed for the analysis
– Birth characteristics
– Anthropometrics through infancy and childhood
– Adult outcomes
– Covariates to adjust for potential confounders
• Establish inclusion criteria
Challenges and solutions: timing of measurements
• Birth: All sites had birth weight, but South Africa and Brazil lacked birth length
• Frequency of child measures varied by site – bi-monthly vs every 3 months, vs annual during the first 2
yr– variable ages at follow-up from 2 yr to adulthood– some sites followed subsamples at specific age
Solutions• Use WHO Z-scores to minimize age differences• Define phases rather than specific ages only
– “mid-childhood”= age 4-5 yr, except for Cebu, where midchildhood = age 8 yr
Challenges and solutions: diverse methods for measuring common outcomes
• Body composition– Estimated from DXA (South Africa)
– equations based on anthropometry (New Delhi, Cebu, Guatemala)
– BIA (Pelotas)
Solutions
– Used site and sex-specific Z-scores of lean and fat mass for analysis
• Glucose– Fasting venous blood
– Fasting capillary/finger stick
– Random non-fasting
Solutions
– Used standard correction factors to harmonize venous and capillary blood, regression analysis using timing of last meal to estimate fasting values for Brazil
Challenges and solutions: common concept, different measures
• Examples include:– Socioeconomic status: Income, wealth, education, social class
• Quality of health care• Maternal autonomy• Urbanicity of residence
Solutions: • Identification of concepts• Inventory of data• Assessment of comparability• Construction of working variables
Concept inventories
Thematic category
Bt20
(South Africa)
Pelotas
(Brazil) Guatemala
CLHNS
(Phillipines)
New Delhi
(India)
Income and SES
Maternal employment status P P P P P
Partner employment status P P
Paternal schooling P P P P P
Maternal schooling P P P P P
Family income P P P P P
SES (measured differentially
across studies, through Grants.
Assets, wealth index) P P P P P
Variable Site N Verbatim Source Question Precise derivation of
composite variable
Child age at
collection
Other child ages at
which equivalent
variable is available
Coding
categories &
description
c3mtscho
Pelotas 5,906 Demography file (number of years of completed education)
N/A Birth 2 Years
years of completed education
Cebu 3,080What is the highest grade have you (mother) completed?
Grade converted to years of schooling
Mother asked duringpregnancy
8 ,11, 15, 18, and 21 yrs
South
Africa2,932
Demography file (number of years of completed education)
Categorical, recoded to continuous using midpoints of each category.No formal educ = 0; gr1-gr5=2.5; gr6-7=6.5; gr8-10=9; gr11-12=11.5; post-high school education=14
0-23, 7, 10, 12, 13, 14, 15, 17, 18 Yrs
Delhi 5,454
Person card (Married Women): 3 – (18) Education (1- Illiterate, 2- Primary, 3-Middle, 4- Matric, 5- College, 6- Literate)
Years of completed education coded as midpoint:Illiterate – 0 yrPrimary – 3 yrsMiddle – 7 yrsMatric – 10 yrsCollege – 12 yrs
Before child birth
Guatemala 2,169 Census file maternal education variables (number of years of completed education)
Birth
COHORTS: Codebook example for creating comparable variables
Challenges and solutions: Which references or standards?
• WHO growth standards for child Z-scores– No controversies
• International Diabetes foundation or WHO definitions for overweight and obesity, hypertension, impaired fasting glucose and diabetes, and central obesity– Proposed alternate cutpoints for Asians?– Youngest participants have low risk
Solutions• Use IDF pre-HTN + HTN, dysglycemia + diabetes• waist to height ratio• WHO BMI cutpoints for weight status definitions
Data analysis
• Meta-regression or pooled data?
• Testing for heterogeneity: site and sex differences
– What to compare: size, sign of coefficients, confidence intervals, statistical significance?
– When to stratify
• Initial COHORTS analysis used metaregression
Adult height according to length/age at 2 yr
1 Z-score at age 2 y = 3.2 cm taller adult
Adjusted for several confounding variablesVictora, Adair et al, Lancet 2008Victora et al. Lancet 2008
Achieved schooling according to length-for-age at 2 yr
Adjusted for several confounding variablesVVictora et al. Lancet 2008
1 Z-score at age 2 y = half a year more schooling
-3 0 3
Mean change in BMI per unit change in birth weight (kg)
Females
Males
Brazil
Brazil
Guatemala
India
Philippines
South Africa
Guatemala
Philippines
South Africa
India
Combined
Meta-analysis of birth weight and adult BMI
VVictora et al. Lancet 2008
Data Analysis: pooled data
• Pooled data from 5 sites: maximizes sample size for analysis (~7500-8000)
• Use regression models to examine associations of early exposures with later outcomes
• Evaluate heterogeneity by site & sex– Include site and sex indicator variables
– Test interactions of key exposures with site and sex
• Evaluate confounding by SES
Systolic Blood Pressure
-3
-2
-1
0
1
2
3
4
5
6
Brazil M
Guatemala M
Delhi M
Cebu M
Soweto M
Brazil F
Guatemala F
Delhi F
Cebu F
Soweto F
pooled
Weight relative to linear growth Linear growth relative to weight gain
*
* Significant sex-site heterogeneity
Birth 24 m MC Adult 24 m MC Adult
mm
Hg
Adair et al Lancet 2013
Challenges and solutions: attrition and missing data
• Attrition rates vary by cohort
• Survey designs led to missing data
– E.g. South Africa and Brazil: not all infants were measured at age 12 mo.
Solution
• Imputation of some missing values
• Inverse probability weighting
brazil
guatemala
delhi
cebu
soweto
brazil
guatemala
delhi
cebu
soweto
brazil
guatemala
delhi
cebu
soweto
brazil
guatemala
delhi
cebu
soweto
All With BW With BW&adult BMIWith BW&adult BMI
&CV4s
site code Freq. Percent Freq. Percent Freq. Percent Freq. Percent
brazil 5,913 26.65 brazil 5,805 29.2 brazil 4,184 41.59 brazil 3,583 47.05
guatemala 2,392 10.78 guatemala 973 4.89 guatemala 552 5.49 guatemala 301 3.95
delhi 7,530 33.94 delhi 6,809 34.25 delhi 1,424 14.15 delhi 1,326 17.41
cebu 3,080 13.88 cebu 3,029 15.23 cebu 2,001 19.89 cebu 1,887 24.78
soweto 3,273 14.75 soweto 3,267 16.43 soweto 1,900 18.88 soweto 518 6.8
Total 22,188 100 19,883 100 10,061 100 7,615 100
COHORTS sample, comparing ALL, those with BW, BW&adult BMI, BW&adult BMI & most
basic set of CVs (CV4, birth, 24 mo, MC, Adult)
Pie charts show site composition across samples
Productivity
1: Addo OY, Stein AD, Fall CH, Gigante DP, Guntupalli AM, Horta BL, Kuzawa CW,Lee N, Norris SA, Osmond C, Prabhakaran P, Richter LM, Sachdev HP, Martorell R;and on Behalf of the Cohorts Group. Parental childhood growth and offspringbirthweight: Pooled analyses from four birth cohorts in low and middle incomecountries. Am J Hum Biol. 2014 Sep 3.
2: Lundeen EA, Stein AD, Adair LS, Behrman JR, Bhargava SK, Dearden KA, GiganteD, Norris SA, Richter LM, Fall CH, Martorell R, Sachdev HS, Victora CG; COHORTSInvestigators. Height-for-age z scores increase despite increasing heightdeficits among children in 5 developing countries. Am J Clin Nutr. 2014Sep;100(3):821-5.
3: Stein AD, Barros FC, Bhargava SK, Hao W, Horta BL, Lee N, Kuzawa CW, MartorellR, Ramji S, Stein A, Richter L; Consortium of Health-Orientated Research inTransitioning Societies (COHORTS) investigators. Birth status, child growth, and adult outcomes in low- and middle-income countries. J Pediatr. 2013Dec;163(6):1740-1746.e4.
4: Adair LS, Fall CH, Osmond C, Stein AD, Martorell R, Ramirez-Zea M, Sachdev HS,Dahly DL, Bas I, Norris SA, Micklesfield L, Hallal P, Victora CG; COHORTS group. Associations of linear growth and relative weight gain during early life withadult health and human capital in countries of low and middle income: findingsfrom five birth cohort studies. Lancet. 2013 Aug 10;382(9891):525-34.
5: Addo OY, Stein AD, Fall CH, Gigante DP, Guntupalli AM, Horta BL, Kuzawa CW,Lee N, Norris SA, Prabhakaran P, Richter LM, Sachdev HS, Martorell R; Consortium on Health Orientated Research in Transitional Societies (COHORTS) Group. Maternalheight and child growth patterns. J Pediatr. 2013 Aug;163(2):549-54.
6: Horta BL, Bas A, Bhargava SK, Fall CH, Feranil A, de Kadt J, Martorell R,Richter LM, Stein AD, Victora CG; COHORTS group. Infant feeding and schoolattainment in five cohorts from low- and middle-income countries. PLoS One. 2013 Aug 20;8(8):e71548.
7: Richter LM, Victora CG, Hallal PC, Adair LS, Bhargava SK, Fall CH, Lee N,Martorell R, Norris SA, Sachdev HS, Stein AD; COHORTS Group. Cohort profile: the consortium of health-orientated research in transitioning societies. Int JEpidemiol. 2012 Jun;41(3):621-6.
8: Kuzawa CW, Hallal PC, Adair L, Bhargava SK, Fall CH, Lee N, Norris SA, Osmond C, Ramirez-Zea M, Sachdev HS, Stein AD, Victora CG; COHORTS Group. Birth weight, postnatal weight gain, and adult body composition in five low and middle incomecountries. Am J Hum Biol. 2012 Jan-Feb;24(1):5-13.
9: Norris SA, Osmond C, Gigante D, Kuzawa CW, Ramakrishnan L, Lee NR, Ramirez-ZeaM, Richter LM, Stein AD, Tandon N, Fall CH; COHORTS Group. Size at birth, weight gain in infancy and childhood, and adult diabetes risk in five low- ormiddle-income country birth cohorts. Diabetes Care. 2012 Jan;35(1):72-9.
10: Fall CH, Borja JB, Osmond C, Richter L, Bhargava SK, Martorell R, Stein AD,Barros FC, Victora CG; COHORTS group. Infant-feeding patterns and cardiovascular risk factors in young adulthood: data from five cohorts in low- and middle-incomecountries. Int J Epidemiol. 2011 Feb;40(1):47-62.
11: Stein AD, Wang M, Martorell R, Norris SA, Adair LS, Bas I, Sachdev HS,Bhargava SK, Fall CH, Gigante DP, Victora CG; Cohorts Group. Growth patterns inearly childhood and final attained stature: data from five birth cohorts fromlow- and middle-income countries. Am J Hum Biol. 2010 May-Jun;22(3):353-9.
12: Martorell R, Horta BL, Adair LS, Stein AD, Richter L, Fall CH, Bhargava SK,Biswas SK, Perez L, Barros FC, Victora CG; Consortium on Health OrientatedResearch in Transitional Societies Group. Weight gain in the first two years oflife is an important predictor of schooling outcomes in pooled analyses from fivebirth cohorts from low- and middle-income countries. J Nutr. 2010Feb;140(2):348-54.
13: Adair LS, Martorell R, Stein AD, Hallal PC, Sachdev HS, Prabhakaran D, Wills AK, Norris SA, Dahly DL, Lee NR, Victora CG. Size at birth, weight gain ininfancy and childhood, and adult blood pressure in 5 low- andmiddle-income-country cohorts: when does weight gain matter? Am J Clin Nutr. 2009May;89(5):1383-92.
14: Victora CG, Adair L, Fall C, Hallal PC, Martorell R, Richter L, Sachdev HS;Maternal and Child Undernutrition Study Group. Maternal and child undernutrition:consequences for adult health and human capital. Lancet. 2008 Jan26;371(9609):340-57.
Pelotas Cohort StudyCesar G. Victora, Pedro C. Hallal, Fernando C. Barros, Bernardo L Horta
and Denise P Gigante(Universidade Federal de Pelotas)
INCAP Nutrition TrialReynaldo Martorell, Aryeh D. Stein (Emory University) Manuel Ramirez-
Zea (Institute of Nutrition of Central America and Panama, Guatemala City )
Cebu Longitudinal Health and Nutrition SurveyLinda S. Adair (UNC Chapel Hill); Judith Borja, Nanette Lee, Isabelita Bas (Office of Population Studies Foundation, University of San Carlos, Cebu,
Philippines); Darren Dahly ( University of Cork); Chris Kuzawa and Thom McDade ( Northwestern University)
New Delhi Birth Cohort StudySantosh K. Bhargava (Sunder Lal Jain Hospital); Harshpal S. Sachdev
(Sitaram Bhartia Institute of Science and Research) ; Caroline Fall, Clive Osmond (MRC Epidemiology Resource Centre,
University of Southampton,UK)
Birth to TwentyLinda Richter ( Human Sciences Research Council, Durban, South Africa)
Shane A. Norris, Lisa Mickleford (Developmental Pathways for Health Research Unit
University of the Witwatersrand, Johannesburg)
Funding for COHORT analysis: Wellcome Trust
Bill and Melinda Gates Foundation