Determinants of infant growth in the slums of Dhaka: size and maturity at birth, breastfeeding and morbidity

Determinants of infant growth in the slums of Dhaka: size and

maturity at birth, breastfeeding and morbidity

SE Arifeen1,2, RE Black2*, LE Caul®eld2, G Antelman1,2 and AH Baqui1,2

1International Centre for Diarrhoeal Disease Research, Bangladesh (ICDDR,B), Dhaka, Bangladesh; and 2Department of InternationalHealth, Johns Hopkins University School of Hygiene and Public Health, Baltimore, Maryland, USA

Objective: To investigate the in¯uences of size at birth, breastfeeding and morbidity on growth during infancy inpoor areas of urban Bangladesh.Design: This was a prospective observational study of a cohort of newborn infants followed until 12 months ofage.Setting: Slum areas of Dhaka City in Bangladesh.Subjects: A total of 1654 newborn infants were enrolled at birth, and follow-up was completed for 1207 infants.Repeated anthropometric measurements and interviews of caretakers on infant feeding and morbidity wereconducted. A mixed effects regression method was used for modeling infant growth.Results: After adjusting for other variables, mean differences in body weight by birth weight and length, small-for-gestational age and prematurity categories remained relatively constant throughout infancy. A positive impactof exclusive breastfeeding in the ®rst 3 ± 5 months on infant growth was detectable at 12 months of age. Althoughthe bigger babies in the sample tended to grow relatively even bigger; exclusive breastfeeding appeared tocounteract this pattern. Reported diarrhoea was associated with lower body weights and lengths even afteradjusting for feeding patterns.Conclusions: Size at birth has an important role in determining growth during infancy. Effective strategies forimproving birth weight, poorly addressed till now in Bangladesh, are needed. The sustained effect on growth andthe even more bene®cial effect in lighter infants are compelling reasons for promotion of exclusive breastfeedingin early infancy.Descriptors: infant growth; birth weight; fetal growth retardation; gestational age; breastfeeding; morbidity;BangladeshEuropean Journal of Clinical Nutrition (2001) 55, 167±178

Introduction

Size at birth is an important determinant of infant growth inboth developed and developing countries (Huttly et al,1990; Adair, 1989; Binkin et al, 1988; Victora et al,1987; Kramer et al, 1985; Garn, 1985; Mata et al, 1975).These studies show that the relative position at birth withrespect to weight and length is maintained in the ®rst yearof life and often throughout childhood, and children withlower birth weights are more likely to remain lighter and

smaller. Nevertheless, the relationship between birthweight and infant growth is in¯uenced by the infant'spropensity for catch-up growth. Newborns, who are bornlight, often demonstrate upward centile crossing of theirbody weights, ie catch-up growth, which usually occurs inthe ®rst year or two of life (Binkin et al, 1988; Peck et al,1987; Persson, 1985).

In developed countries, catch-up growth is most notice-able among preterms (Binkin et al, 1988; Piekkala et al,1989; Cruise, 1973). This phenomenon is now consideredlargely the result of relating the growth rate of prematureinfants to the chronological age since birth rather than agesince conception (Casey et al, 1990; Karniski et al, 1987).Although small-for-gestational age (SGA) infants showsome growth spurts in the ®rst 3 ± 6 months of life,especially those with asymmetric growth retardation, theyremain persistently shorter (Castillo-Duran et al, 1995;Tenovuo et al, 1987; Villar et al, 1984).

Data on these relationships in developing countries arelimited. In Bangladesh, the prevalence of low birth weight(LBW) is very high, as is child malnutrition. A betterunderstanding of the role of nutritional status at birth in

*Correspondence: RE Black, Department of International Health, JohnsHopkins University School of Hygiene and Public Health, 615 N WolfeStreet, Baltimore, MD 21205, USA.E-mail: [email protected]: SE Arifeen.Contributors: SEA, REB, GA and AHB developed the study and itsdesign. SEA and GA were responsible for the implementation of ®eldwork. SEA, REB and LEC designed the analysis. SEA drafted themanuscript, which was then revised by all the investigators.Received 27 March 2000; revised 26 October 2000; accepted31 October 2000

European Journal of Clinical Nutrition (2001) 55, 167±178ß 2001 Nature Publishing Group All rights reserved 0954±3007/01 $15.00

www.nature.com/ejcn

infant growth would facilitate appropriate policy and pro-grammatic decisions. This should also involve considera-tion of the simultaneous effects of infant feeding andmorbidity on growth. There is considerable evidence of agrowth promoting role of exclusive (Mahmood & Feacham,1987; Fawzi et al, 1997) or predominant (full) breastfeed-ing (Adair et al, 1993; Martines et al, 1994) in the ®rst 6months of life. Available evidence also suggests that acuterespiratory infections (ARI) (Adair et al, 1993; Smith et al,1991; Cole, 1989; Rowland et al, 1988) and diarrhoea(Adair et al, 1993; Checkley et al, 1997; Bittencourt et al,1993; Heikens et al, 1993; Becker et al, 1991; Zumrawiet al, 1987; Black et al, 1984) have negative effects ongrowth.

This study was carried out in a sample of the slumpopulation of Dhaka, Bangladesh to assess the effect ofnewborn size and maturity de®ned by weight, gestationalage and gestational age-speci®c weight on infant growth,and to examine how other factors, especially feedingpractices and morbidity, in¯uenced growth during infancy.

Methods

A total of 1654 singleton, live born newborns were enrolledin the study during 1993 ± 1995 from a sample of slumareas of ®ve thanas (administrative areas) of Dhaka City.Enrollment was limited to newborns who could be reachedwithin 13 days of birth. A total of 181 newborns were notenrolled as they were reached within �14 days after birth.Each baby's weight and length were measured at enroll-ment and information was collected on feeding and illnesssince birth. A standardized postnatal physical gestationalage assessment was conducted (Capurro et al, 1978).Information on household and parental characteristics, themother's fertility experience and the date of her lastmenstrual period (LMP) were collected before birth.Follow-up visits were planned at 1, 3, 6, 9 and 12 monthsof age. At each visit, weights were measured and structuredquestionnaires were used to collect information on morbid-ity and feeding. Follow-up was not completed for 392(24%) of the study cohort because of death (157) oroutmigration (235).

VariablesBody weight and length were selected as the dependentvariables in this analysis. A maximum of six weight andlength measurements were available for each child, includ-ing measures at enrollment. The explanatory variables canbe categorized into two groups: (i) the key variables ofinterest, ie classi®cation of size and maturity status at birth,and feeding and morbidity status; and (ii) the measuredconfounding variables.

Classi®cation of size and maturity status at birth. New-borns were classi®ed based on their estimated birth weight,gestational age and proportionality. Birth weight was esti-mated from weights at enrollment. Because babies were

enrolled as many as 13 days after birth, some adjustmentwas necessary in order to characterize status at birth. In asub-sample of 99 newborns daily repeat weights wereobtained for 14 days after birth, thus permiting an analysisof postnatal changes in body weight in this population(Arifeen, 1997). This analysis revealed that there was verylittle change in body weights in the ®rst 72 h; and thusenrollment weights obtained < 72 h after birth were usedfor birth weights for about 75% of the infants. For theremaining infants who were enrolled 3 ± 13 days after birth,birth weights were estimated from enrollment weightsusing a regression equation developed from the sub-sample with daily repeat weights. The number of com-pleted weeks between the date of LMP and date of birthwere used as the gestational ages of most newborns.However, for 246 newborns (15%) the LMP was missingor considered invalid. The Capurro physical assessmentwas used to estimate gestational age of these infants.

Three sets of variables were created: (i) a two-level birthweight variable (< 2500 g vs �2500 g); (ii) a four-categoryvariable (named SGA=GA) indicating SGA, ie birth weightless than the 10th percentile of a US reference popula-tion (Alexander et al, 1996) or AGA (appropriate-for-gestational age) and preterm, ie < 37 weeks of gestationor term; and (iii) a three-category variable (named SGAtype) indicating SGA=AGA status and proportionality ofSGA infants, ie low ponderal index (PI) if the index (weight(g)=length (cm)3) was less than the 10th percentile on areference chart of adequate PI (Lubchenco et al, 1966).

Feeding. At each visit, infant feeding practices in theprevious 7 days were classi®ed as exclusively breastfed(only breast milk); predominantly breastfed (breast milkwith water, sugar water, honey, or other non-milk liquid);partially breastfed (breast milk with animal, powdered orcondensed milk, cereals, eggs, vegetables, fruits, lentils,meat, or ®sh); and not breastfed. Feeding information fromthe visits scheduled for months 1 and 3 was used to create asummary measure of breastfeeding status in the ®rst 3months of life. Infants exclusively breastfed at both visitswere classi®ed as exclusively breastfed. Infants exclusivelyor predominantly breastfed at each visit, but not exclusivelybreastfed at both visits were classi®ed as predominantlybreastfed. The remaining infants were grouped together andmainly included infants who were partially breastfed at oneor both visits. When information from one visit wasmissing, data from the other visit was used (8.4%), andwhen both visits had missing data, status at enrollment wasused (1%). Information at enrollment was otherwiseexcluded since most newborns were given other liquidsfor a few days after birth and feeding status at this age didnot re¯ect the overall feeding during the ®rst 3 months oflife. Another variable indicating breastfeeding status at 6months was included in selected analyses.

A measure of complementary feeding was created byusing a count of the different types of complementarydrinks or foods given to infants in the 7 days precedingthe visit (Marquis et al, 1997). There were nine possible

Determinants of infant growth in Dhaka slumsSE Arifeen et al

168

European Journal of Clinical Nutrition

food=drink items for this measure: sugar water, honey, milk(non-breastmilk), cereal, egg, vegetables, fruits, pulses, andmeat ± ®sh.

Morbidity. Two-week prevalences of diarrhoea and ARIwere included in the analysis as measures of infant mor-bidity. The 2-week period preceding the visit to the studysubject was considered positive for diarrhoea if frequentloose or liquid stools were reported with peak stool fre-quency of three or more and with at least two of thefollowing symptoms present: weakness, thirst, dry mouth,loose skin, depressed fontanelle, sunken eyes, and no orlittle urine. ARI was considered present in the 2-weekperiod if the caretaker reported three or more of thefollowing symptoms including at least one of the under-lined symptoms: cough, dif®cult breathing, noise duringbreathing, nasal ¯aring, blue lips=skin, chest indrawing,rapid breathing, not feeding well or not able to drink, andabnormally sleepy or dif®cult to wake. This was adaptedfrom diagnostic algorithms used in other studies (Baquiet al, 1998).

Other variables. The control variables in the analysisinclude both proximate determinants of growth such aschild's age and sex; and other parental and householdvariables that act either through the measured proximatedeterminants or other unmeasured factors. The parental andhousehold variables were included in the regression modelsto account for the unmeasured factors.

The study was approved by the Ethical Review Com-mittee of ICDDR,B and the Committee on HumanResearch of the Johns Hopkins University School ofHygiene and Public Health. All data was obtained afterinformed verbal consent from at least one parent=caretakerof the infant.

StatisticsDescriptive statistics of the sample were estimated. Mixedeffects regression analysis was performed. These regressionmodels had two components: random effects parameters, ierandom intercepts, and random linear and curved slopes foreach infant; and ®xed effects parameters (Diggle et al,1994). Allowing each child to have its own intercept andslope accounted for the correlation between multiple mea-sures on the same individual. The ®xed effects part of themodels was built in two stages. In the ®rst stage only theconfounding variables were included in a block and onlythose variables with overall P-values less than 0.10 wereretained. Age was modeled with three continuous variablesincluding age, age3 and a cubic spline with a knot at 3months. The cubic spline was used to describe the smoothaverage growth curve. The position of the knot wasdetermined by visual inspection of scatter plots by ageand plots of mean growth curves. The weights and lengthsat enrollment contributed to the estimation of the randomintercepts, which corresponds to the measurements at birth.Because age at measurement was an explanatory variable inthe model, the use of estimated birth weights rather than

actual weights at enrollment was unnecessary. In thesecond stage, ®xed effects parameters representing thekey explanatory variables of interest were added, ie eitherbirth weight category, SGA=GA category or SGA type, allbased on estimated birth weight; and feeding and morbiditystatus. Interaction terms with age were also included. Foreach of the three different sets of variables, separate modelswere developed for the two age strata: the four measure-ments from 0 to<6.5 months, and the three measurementsfrom 6 months onwards. Choosing 0 ± 6.5 months for theyounger set was necessary to include most of the measure-ments at age 6 months so that the model had enough datapoints at the upper extreme. The ®nal models only includedvariables still signi®cant at the 0.1 level. An additionalvariable indicating breastfeeding status at 6 months wasincluded in the models with older infants but was droppedas it was not signi®cant. Additional models were developedwithout any of the three size and maturity variables toassess the relationship between random intercepts andslopes for different feeding categories.

Since poorer growth is associated with increased like-lihood of weaning or changing to partial breastfeedinginstead of exclusive breastfeeding, such `reverse causality'tends to overestimate the effect of breastfeeding on growth(Victora et al, 1998). Since models with random effects donot completely resolve this problem we decided to speci-®cally assess this. Attained weight=length was regressed onweight=length and breastfeeding practice at the previousvisit in models with only ®xed effects. No random effectswere included in these models. This methodology had beenused by Victora et al (1998).

All analysis was performed with the SAS System soft-ware for Windows (version 6.12) of the SAS Institute,Cary, NC, USA. For the mixed effects regression analysis,the SAS procedure PROC MIXED was used, with therestricted maximum likelihood method for computation.

Results

Almost three-quarters of the 1654 sample of newbornswere enrolled within 72 h of birth. The mean of theestimated birth weights was 2516 g and almost half of thenewborns (46.4%) were low-birth-weight (< 2500 g). Pre-term deliveries accounted for 17% of all infants, and almost70% of the sample were SGA. Infants with adequateponderal index (PI) comprised more than two-thirds of allthe SGA infants.

Boys and girls numbered almost equally in the sample.The average age of their mothers was about 25 y. Only 25%of the mothers had received any formal education, 95%were Muslims, slightly less than a third of the mothers wereborn in Dhaka, 62% were shorter than 150 cm, and 31%gave a history of previous child death. About a quarter ofthe index children were ®rst borns. The mean monthlyincome of the child's family was about US$75 and almost40% of the families did not have a TV, radio, bicycle,rickshaw, cupboard, table, watch or a clock.


169


A little more than half of the infants were exclusivelybreastfed at 1 month of age, declining to about a quarter at3 months (Table 1). Widespread use of pre-lacteal feedsexplains the low prevalence of exclusive breastfeeding atenrollment. According to the summary measure for overallbreastfeeding status in the ®rst 3 months of life, 24.1%were exclusively breastfed, and 16.6% predominantly

breastfed. Almost all of the others were partially breastfed.The mean number of types of complementary drinks andfoods given at enrollment was 1.7, out of the maximumpossible of 9. The mean declined to less than 1 at ages 1and 3 months. Thereafter, the number of different types ofcomplementary drinks and foods given to infants increasedsteadily with age to reach a mean of 3 at 12 months. The2-week prevalence of diarrhoea steadily increased withage, reaching about 26% at 12 months. ARI prevalencewas more stable across ages at about 13.5 ± 18.4%.

The focus in the ®rst 12 mixed effects regression modelswas on estimating the overall ®xed effects of the explana-tory variables of interest (Tables 2 and 3). Table 2 presentsthe ®xed effects estimates of the four models with birthweight categories, and Table 3 presents the same forselected variables in the models with the SGA=GA andSGA-type variables.

Being male, or the child of an educated or a tallermother, was strongly associated with higher achievedweights and lengths (Table 2). In early infancy, youngerand lower parity mothers had lighter babies, but this

Table 1 Percentage distribution of enrolled infants by age and currentbreastfeeding practicesa

Breastfeeding status

Age in months n Exclusive Predominant Partial� none

0 (� 15 days)b 1654 6.2 65.9 27.91 (� 15 days) 1469 53.1 14.4 32.53 (� 15 days) 1143 25.5 14.0 60.56 (� 15 days) 1090 4.8 10.9 84.39 (� 15 days) 1077 0.8 5.5 93.7

12 (� 15 days) 1078 0.0 4.1 95.9

aBased on breastfeeding practices in the 7 days preceding each visit.bFigures in parentheses indicate range of acceptable visit dates.

Table 2 Results (coef®cients) of multiple regression analysis of body weight (g) and length (cm) using mixed models:a 2 ®xed effects for breastfeeding,morbidity, birth weight, and other covariates

Weight Length

0 ± 6.5 months � 6 months 0 ± 6.5 months � 6 monthsFixed effects parameters only Model 1 Model 2 b Model 3 Model 4 b

Intercept 2045.2§ 6341.4§ 47.10§ 64.47§

Age 945.8§ 177.0§ 4.20§ 1.11§

Age3 7 14.2§ 7 0.7§ 7 0.08§ 7 0.003§

[(Age-3)�]3 41.4§ Ð 0.26§ ÐNormal birth weight (reference: LBW) 603.4§ 547.3§ 2.60§ 1.71§

Normal birth weight and age (reference: LBW) 13.8* NS 7 0.06{ NSBreastfeedingc (reference: exclusive) Predominant NS 117.1 NS 0.25

Partial� none 7 224.8} 7 0.54}

Breastfeedingc and age (reference: exclusive) Predominant 15.6 7 31.7{ 0.07* NSPartial� none 7 36.7} 21.7{ 7 0.06{

Complementary foods and drinksd (no of types) 7 25.7§ NS 7 0.14§ NSNo diarrhoea (reference: diarrhoea present) 172.4§ 153.6§ 0.28} 0.17{

No ARI (reference: ARI present) 32.2{ 105.3} NS 7 0.12*Infant is a girl (reference: being a boy) 7 33.0{ 7 467.2§ 7 0.61§ 7 0.58§

Mother's age (y) (reference: 30� ): 13 ± 19 7 65.6{ 7 64.8 7 0.39{ 7 0.39{

20 ± 29 7 22.9 177.7{ 7 0.15 7 0.16Mother has no schooling (reference: any schooling) 7 44.3{ 7 204.9} 7 0.26{ 7 0.24{

Muslim household (reference: other religion) 64.2{ NS 0.44{ 0.39{

Mother born in Dhaka (reference: born elsewhere) 40.4{ NS NS NSMonthly income (US$) (reference: 75� ): 0 ± 49 NS 7 153.6{ NS NS

50 ± 74 7 64.2Level of household assets (reference: high): Low NS 7 204.8{ 7 0.32{ 7 0.30{

Medium 7 22.5 7 0.19* 7 0.19*Number of previous live births (reference: 3� ): 0 7 65.7{ 225.5{ NS NS

1 ± 2 7 10.0 85.5Mother's height (cm) (reference: 150.0 ± 168.0): 126.3-144.9 7 55.2{ 7 388.0§ 7 0.77§ 7 0.75§

145.0-149.9 7 44.7{ 7 187.5} 7 0.50§ 7 0.48§

Signi®cance based on t-tests: *0.10; {< 0.05; {� 0.01; }� 0.001; §� 0.0001.aMixed models include both ®xed and random effects. The table only presents the results of the ®xed effects. The random parameters included in the modelsare (results not shown): intercept, age, age3 (models 1, 3) and intercept, age (models 2, 4). The random parameter for age3 was dropped in models 2 and 4 dueto linear dependencies with the random parameter for linear age.bModels 2 and 4 on the 6� sub-sample uses age and age3 rescaled to start at 0 from 6 monthscSummary measure of breastfeeding status in the ®rst 3 months of life.dComplementary foods and drinks include: sugar water, honey, milk (non-breastmilk), cereal, egg, vegetables, fruits, pulses, meat ± ®sh.


170


association reversed with increasing infant age. Parity wasnot associated with length while younger mothers hadshorter infants at any age. There was a strong positiveassociation between attained weight of the infants at olderages and household economic status as measured byincome and possession of assets. However, only householdassets predicted lengths of the infants. Infants with priorsibling death also weighed less in comparison with infantshaving no such history, but there was no association withlength. Religion was associated with both weight andlength and the mother's place of birth was only associatedwith body weight.

On average, there was about 600 g mean difference inweights and about 2 cm difference in lengths of normal andlow-birth-weight babies during infancy, with the differencebeing smaller in the second 6 months (Table 2). There wasa signi®cant effect of birth weight on the age slope forlength from 0 to 6.5 months, the gain in length with agebeing slower for normal birthweight (NBW) infants. Theeffect of birth weight on the age-slope for weight in theyounger ages was in the reverse direction though notstatistically signi®cant.

The effect of classi®cation by AGA=SGA and term=preterm status was very strong (Table 3). Compared topreterm-SGA infants, term-AGA infants were about 1 kgheavier and 3.1 ± 5.6 cm taller, while term-SGA andpreterm-AGA infants were intermediate in weight andlength, although their relative positions in comparison topreterm-SGA infants were reversed in the second half ofinfancy. SGA infants with low PI were lighter but tallerthan those with adequate PI, while AGA infants were bothheavier and taller than the SGA infants. Overall, thedifference in weight and length between the variousgroups was narrower in the older infants. There wassigni®cant interaction between age and the SGA=GA andSGA type variables in the young infant models for length,indicating a slower gain in length among infants born taller.

In the young infant models (0 ± 6.5 months), the maineffect for breastfeeding was not signi®cant, that is, therewere no differences in the intercept (baseline) by breast-feeding mode (Table 2). The signi®cant breastfeeding ± ageinteraction term means that the infants differ in their slopes(growth rate) by breastfeeding mode. In this age stratum,compared to the exclusively breastfed (EBF), the partiallyor not breastfed infants gained weight and length at asigni®cantly slower rate. The predominantly breastfed(PBF) infants gained most rapidly but the difference withEBF was small and non-signi®cant. In the older ages(6�months), the signi®cant main effect was a consequenceof the diverging slopes in the younger ages (models 1 and3). The interaction term between breastfeeding mode andage was signi®cant for the weight model (model 2),indicating further change in growth rates for weight. Ofinterest was the reversal of the slopes in this age group.PBF infants, who start at the highest level at 6 months, gainweight at a slower rate than the EBF infants, whereas thepartially or not breastfed infants had the fastest growth rateafter being at the lowest level at 6 months. Such an effectwas not observed for length.

The negative effects of diarrhoea in the preceding 2weeks were strong and signi®cant for both weights andlengths and the effects were similar in both age groups.ARI was signi®cantly associated with weight but notlength, and the size of the effect was larger in older infants.

Babies partially or not breastfed at month 1 were 0.5 cmshorter and about 100 g lighter at 3 months of age comparedto those who were exclusively breastfed (Table 4). Inclu-sion in the models of weight or length at the preceding visitand other covariates resulted in substantial reduction in thesize of the effect of breastfeeding at the previous visit ongrowth at the following visit. There were signi®cant un-adjusted effects in the 0 ± 1 and 3 ± 6 month intervals, butthese effects disappeared when the other variables areincluded in the models.

Table 3 Results (coef®cients) of multiple regression analysis of body weight (g) and length (cm) using mixed models: ®xed effects for SGA=GA and SGAtypea

Weight Length

Fixed effects parameters only 0 ± 6.5 months � 6 months 0 ± 6.5 months � 6 months

Model 5 Model 6 Model 7 Model 8SGA=GA (reference: pre-term SGA): Term AGA 1183.4§ 1000.3§ 5.56§ 3.14§

Pre-term AGA 653.4§ 442.8{ 3.43§ 1.56}

Term SGA 601.5§ 485.9{ 3.35§ 1.60§

SGA=GA and age (reference: pre-term SGA): Term AGA NS NS 7 0.27} NSPre-term AGA 7 0.18{

Term SGA 7 0.16{

Model 9 Model 10 Model 11 Model 12SGA type (reference: SGA-adequate PI): AGA 317.8§ 249.4§ 1.71§ 1.07§

SGA-low PI 7 184.6§ 7 152.9{ 0.70§ 0.32*SGA type and age (reference SGA-adequate PI): AGA NS NS 7 0.10} NS

SGA-low PI 7 0.05*

Signi®cance based on t-tests: *0.10; {< 0.05; {� 0.01; }� 0.001; §� 0.0001.aOnly coef®cients for SGA=GA and SGA type have been presented.


171


Separate models were constructed for the two age stratain which the random effects were either allowed or notallowed to vary by breastfeeding status (Tables 5 and 6).These models were designed to assess the relationshipbetween the random parameters and the in¯uence ofbreastfeeding on these associations. In mixed effectsmodels, the mean and spread of the random interceptswould be expected to closely approximate that of theactual birth weights and lengths, if the latter were astrong determinant of subsequent growth. This was evidentfrom the model results. According to model 13, the meanweight at age 0 (the intercept) was 2475 g, just 41 g lessthan the actual mean birth weight. The standard deviationof the random intercepts was about 375 g, compared to404 g of the actual birth weights. The model was thuspredicting a birth weight distribution that was quite close tothe actual. The correlation between the two random para-meters was indicative of the type and strength of therelationship between the two. Unlike analysis based onmain-effects-slope interactions, which only look at the

average effects, each child has his or her own interceptand slope in these models, and the relationship betweenthem is of interest.

To simplify interpretation, we ignored the randomcurvature (age3) in these models, and only modeled arandom linear slope (age) for the younger age range(< 6.5 months). It was assumed that within the narrowage range the random curvature was negligible, althoughstatistically signi®cant, and ®xed effects parameters onlymay be an adequate model of the growth curve from 0 to6.5 months.

The overall random effects in model 13 demonstrate thatduring the early months, weight at birth was positivelycorrelated with growth rate, ie the heavier the babies wereat birth, the faster they grew (Table 5). The correlation wasnot very high (0.15), suggesting considerable variability ingrowth rates relative to initial size. The correlation remainsabout the same (0.11 ± 0.24) across all three breastfeedingcategories (model 14). Although the random interceptstandard deviation, which represents variability in birthweights, also remains the same irrespective of the breast-feeding status, the random slope (growth rate) standarddeviation was greater in the partial or not breastfed cate-gory suggesting greater variability of growth rates in thisgroup of infants, possibly re¯ecting, in part, the greatervariability of feeding practices. The correlation betweenrandom intercepts and slopes in the exclusively breastfedinfants, although positive, was not signi®cant, and impliesthat in these infants, postnatal growth rate was not asso-ciated with birth weight. After 6 months the variability inthe growth rates of individual infants was less (models 15and 16). As a consequence of the greater variability in thegrowth rates of partially or not breastfed infants in theearlier months, the random intercepts (at 6 months) forthese infants were much more variable. This same group ofinfants exhibited the least variable growth rates after 6months. The overall intercept ± age correlation was higherand still positive, ie the heavier the babies were at 6 monthsthe faster they grew; however, this relation was onlystatistically signi®cant in the partially or not breastfedinfants, in whom the correlation coef®cient was as highas 0.50. Predominantly breastfed infants had a positive butnon-signi®cant intercept ± age correlation whereas, amongexclusively breastfed infants, this parameter was negativebut not signi®cant, suggesting possible catch-up growth, iesmaller babies growing faster.

The ®ndings with the same analyses with length weresomewhat different (Table 6). Based on the overall randomeffects in model 17, length at birth was negatively asso-ciated with growth rate in the ®rst 6 months of life,although, as with weight, the correlation was not high(0.17). Strati®cation by breastfeeding status shows highernegative correlation when the child was exclusively orpredominantly breastfed, and non-signi®cant correlationfor partially or not breastfed babies. When the infantswere older, there was less variability in the length growthrates of individual infants (models 19 and 20). Possibly as aconsequence of the `catch-up' growth seen in exclusively

Table 4 Effect of breastfeeding practices on growth: results from ®xedeffects models

Regression coef®cients

Breastfeeding practices and growth Unadjusted a Adjusted b

Breastfeeding practices at month 0 and growth from months 0 to 1Weight (g):

Exclusive 0 0Predominant 7 66 7 3Partial� none 7 116* 7 50

Length (cm):Exclusive 0.0 0.0Predominant 7 0.5* 7 0.2Partial� none 7 0.5* 7 0.3


Exclusive 0 0Predominant 20 7 26Partial� none 7 276§ 7 103}

Length (cm):Exclusive 0.0 0.0Predominant 0.2 7 0.2Partial� none 7 0.5{ 7 0.5}


Exclusive 0Predominant 113 52Partial� none 7 239} 19

Length (cm):Exclusive 0.0 0.0Predominant 0.3 0.3Partial� none 7 0.3 7 0.0

Signi®cance based on t-tests: *0.10; {< 0.05; {� 0.01; }� 0.001;§� 0.0001.aYx�BF(x71), where Yx�weight=length at time x (eg month 3);BF(x71)� breastfeeding at time x7 1 (eg month 1).bYx�Y(x71)�BF(x71)� covariates where Y(x71)�weight=length at timex7 1 (eg month 1); covariates� child's age at time x; days from timex7 1 to x and sex; 2-week prevalence of diarrhoea and ARI at time x;mother's age, education and height; and household assets and religion.


172


or predominantly breastfed infants in the ®rst 6 months, thevariability in the random intercepts (at 6 months) for theseinfants was much lower than in those who were partially ornot breastfed. These infants also had the least variablegrowth rates after 6 months. The overall intercept-agecorrelation was positive in the older ages, ie the taller thebabies were at 6 months the faster they grew; however,this relationship was no longer statistically signi®cantwhen strati®ed by breastfeeding status, although it wasconsistently positive.

Discussion

Growth during infancy was strongly in¯uenced by size andmaturity at birth in this population. This is consistent with®ndings from other studies in widely different populations(Huttly et al, 1990; Peck et al, 1987; Adair et al, 1993;Dewey, 1998). Dhaka slum infants appear to be con®nedwithin the `growth channels' determined at birth (Arifeenet al, 2000). Mean differences in weights and lengthsbetween categories de®ned at birth were similar in bothhalves of infancy, even after accounting for the effects ofother factors in the models. In contrast to ®ndings in otherdeveloped and developing country populations (Adair,

1989; Binkin et al, 1988; Cruise, 1973), pretermand SGA-low PI infants in our study did not demon-strate faster growth rates. There was some indication thatpreterm-AGA infants gained length compared to theterm-AGA infants in the ®rst 6 months, but they stillremained 1.6 cm shorter at 6 months of age and there wasno further gain. SGA-low-PI infants actually gained lengthat a slower rate in the ®rst 6 months of life comparedto SGA-adequate-PI babies. Even though there was nosigni®cant interaction between age and SGA=GA forweight, the difference between term-SGA and preterm-SGA and between preterm-AGA and preterm-SGA infantsdeclined from more than 600 g in the ®rst 6 months to morethan 400 g in the second half of infancy. The greatermortality among the preterm-SGA (25%) may partiallyexplain this ®nding. In the models for length, the negativeinteraction between age and SGA=GA largely explains thenarrowing in attained length seen at 6 months. For example,the mean difference in length between term-AGA andpreterm-SGA infants declined by 2.42 cm from 5.56 atbirth to 3.14 cm at 6 months. Two-thirds of this reductionwas explained by the 0.27 cm decline in the age slope permonth.

According to the timing hypothesis for the etiology ofintrauterine growth retardation (Caul®eld et al, 1991; Villar

Table 5 Results (coef®cients) of multiple regression analysis of body weight in grams using mixed models (®xed andrandom effects): overall random effects and random effects strati®ed by breastfeeding status

0 ± 6.5 months � 6 months

Parameters Model 13 Model 14 Model 15a Model 16a

Random effects parametersOverall:

Intercept (standard deviation) 375.10§ Ð 775.94§ ÐAge (standard deviation) 141.65§ Ð 57.82} ÐIntercept� age (correlation coef®cient) 0.15§ Ð 0.23{ Ð

By breastfeeding statusExclusive:

Intercept (standard deviation) Ð 389.69§ Ð 738.56§

Age (standard deviation) Ð 118.20§ Ð 85.23§

Intercept� age (correlation coef®cient) Ð 0.11 Ð 7 0.14

Predominant:Intercept (standard deviation) Ð 376.52§ Ð 644.61§

Age (standard deviation) Ð 110.22§ Ð 68.10{

Intercept� age (correlation coef®cient) Ð 0.24{ Ð 0.17

Partial� none:Intercept (standard deviation) Ð 368.71§ Ð 819.33§

Age (standard deviation) Ð 156.33§ Ð 41.12*Intercept� age (correlation coef®cient) Ð 0.15} Ð 0.54§

Residual standard deviation 239.88§ 239.75§ 432.59§ 431.36§

Fixed effects (age parameters only):b

Intercept 2475.0§ 2469.8§ 6836.3§ 6858.9§

Age 940.3§ 943.2§ 177.3§ 176.6§

Age3 7 12.7§ 7 12.8§ 7 0.8§ 7 0.7§

(Age-3)3 [0 if age� 3] 32.7§ 33.2§ Ð Ð

Signi®cance of parameters are based on type III F-tests for ®xed effects; and Wald Z-tests for random effects: *0.10;{� 0.05; {� 0.01; }� 0.001; §� 0.0001.aModels 15 ± 16 on the � 6 month sub-sample uses age and age3 rescaled to start at 0 from 6 months.bOther ®xed effects covariates: as in models 1 and 2 (Table 1).


173


& Belizan, 1982; Villar et al, 1989; Urrusti et al, 1972),SGA-adequate-PI infants are the result of chronic intrau-terine stress, ie since early in pregnancy. This, on the onehand, may make them more adapted to the extrauterineenvironment, but on the other hand, may result in areduction in their capacity for postnatal growth (Villaret al, 1989; Rosso, 1989). SGA-low-PI infants, born withadequate length and brain size, probably result from intra-uterine stress late in pregnancy and thus, it is said, morelikely to grow well after birth (Villar et al, 1989). Such aphenomenon was not clearly seen in our study. Because ofthe high prevalence of fetal growth retardation and theconsiderable shift to the left of the gestational age-speci®cbirth weight distribution, we believe that most of the infantsin this cohort would have experienced some degree of fetalgrowth retardation which resulted in a reduction in theirpropensity for growth after birth. This probably explainsthe observed lack of `catch-up' growth among the SGA-low-PI as well as the limited `catch-up' growth among thepreterm infants.

The mixed effects regression analysis partially over-comes the limitation of an analysis based on averagechanges which hide the variability between individuals

with some infants exhibiting catch-up and others showingsevere faltering. The evidence presented here on the basisof such analysis lends strong support to the hypothesis thateach infant has his or her own growth curve, the level andcourse of which is determined largely by size at birth. Inaddition, there was signi®cant variation between the growthrates and the curvature of the growth curves of the infants.The much greater variation in the random slopes (growthrates) in the younger ages compared to the later half ofinfancy reinforces the argument for targeting growth pro-motion interventions at these early ages (Schroeder et al,1995; Lutter et al, 1990). The overall growth pattern in thispopulation appears to be that of diverging growth curveswith the heavier infants getting even heavier relative tothose who start at lower levels, although the shorter babiesdo appear to make up some of the shortfall in the ®rst 6months, but in the second half of infancy we again observethat taller babies grew relatively even taller. This is unlikethe situation seen in developed countries where the weightsof children converge towards the mean for the ®rst year ortwo and then stabilize into respective tracts (Binkin et al,1988; Persson, 1985). It would be logical to presume that,in a population where most infants are born with certain

Table 6 Results (coef®cients) of multiple regression analysis of body length in cm using mixed models (®xed andrandom effects): overall random effects and random effects strati®ed by breastfeeding status

0 ± 6.5 months � 6 months

Parameters Model 17 Model 18 Model 19a Model 20a

Random effects parametersOverall:

Intercept (standard deviation) 1.98§ 2.23§

Age (standard deviation) 0.31§ 0.19§

Intercept� age (correlation coef®cient) 7 0.17} 0.13{

By breastfeeding statusExclusive:

Intercept (standard deviation) 1.91§ 1.90§

Age (standard deviation) 0.24§ 0.14{

Intercept� age (correlation coef®cient) 7 0.33} 0.16

Predominant:Intercept (standard deviation) 2.13§ 1.91§

Age (standard deviation) 0.17{ 0.07Intercept� age (correlation coef®cient) 7 0.44{ 0.17

Partial� none:Intercept (standard deviation) 1.97§ 2.42§

Age (standard deviation) 0.36§ 0.23§

Intercept� age (correlation coef®cient) 7 0.10* 0.13*

Residual standard deviation 1.21§ 1.20§ 1.03§ 1.04§

Fixed effects (age parameters only):b

Intercept 49.13§ 49.18§ 64.34§ 64.35§

Age 4.15§ 4.15§ 1.11§ 1.11§

Age3 7 0.07§ 7 0.07§ 7 0.003§ 7 0.003§

(Age-3)3 [0 if age� 3] 0.25§ 0.24§

Signi®cance of parameters are based on type III F-tests for ®xed effects; and Wald Z-tests for random effects: *0.10;{� 0.05; {� 0.01; }� 0.001; §� 0.0001.aModels 19 ± 20 on the � 6 month sub-sample uses age and age3 rescaled to start at 0 from 6 months.bOther ®xed effects covariates: as in models 3 and 4 (Table 1).


174


degrees of fetal growth retardation, the biggest would bethe least growth retarded and thus would be more likely tohave preserved their capacity for growth.

The study has recon®rmed the bene®cial effect ofbreastfeeding on growth even after accounting for otherfactors. Several studies have presented evidence of anassociation between birth weight and subsequent feedingpractices with mothers of babies who are smaller at birthbeing less likely to initiate breastfeeding, or if breastfeed-ing is started, more likely to terminate breastfeeding earlier(Barros et al, 1986; Adair & Popkin, 1996). Such arelationship, if present, would explain some of theobserved association between feeding practices andgrowth. Data not presented showed that the proportionexclusively breastfed (EBF), partially breastfed (PBF) orpartially=not breastfed did not differ at any age betweenNBW and LBW babies. The lack of signi®cant maineffects in the models 1 and 3 in Table 2 also indicatesthat there was no association between birth weight orlength and breastfeeding mode in this sample. Studieshave also shown that infants who grow poorly are morelikely to have either their breastfeeding stopped or supple-mented with other foods (Victora et al, 1998). Such reversecausality, if not taken care of in the analysis, may over-estimate the bene®cial effect of breastfeeding. One way ofdoing that is to regress weight at the end of an age intervalon weight and breastfeeding at the beginning of thatinterval. We saw in Table 4 that, compared to exclusivebreastfeeding, partial or no breastfeeding at month 1 wasnegatively associated with weight and length at month 3,even after adjusting for weight=length at month 1 and otherconfounders. The reduction in the breastfeeding effectsafter adjustment suggests though that there is some `rever-sal causality' effect, and indicates that the unadjustedmeasures were overestimating the actual effects. Otherpossible explanations for the lack of an effect after adjust-ment in the 0 ± 1 and 3 ± 6 month intervals may include thefact that very few babies were being exclusively breastfedat enrollment and most of those who were being exclu-sively breastfed at 3 months were no longer fed that way at6 months of age and in many cases this change may haveoccurred soon after the month 3 visit. These ®xed effectsmodels have the limitation of only using information ongrowth at two points in the child's life, while the mixedeffects models uses all the information from the child aswell as from other children in the sample, and thusprovides better estimates. Although the mixed effectsmodels are limited in their ability to handle the bias dueto reverse causality, the consistency of the ®ndings withthose from the ®xed effects models in Table 4 is reassur-ing. The use of a summary measure of breastfeeding in ouranalysis also reduces the likelihood of later weight orlength in¯uencing the measurements.

Breastfeeding is known to in¯uence growth throughseveral pathways, especially the provision of nutrientsand reduction of morbidity such as diarrhoea (Adair et al,1993; Martines et al, 1994; Feacham & Koblinsky, 1984).The signi®cant and strong effect of breastfeeding, even

after adjusting for diarrhoea and ARI morbidity, wasimpressive and is indicative of the nutritional bene®ts ofbreastfeeding (as opposed to the morbidity prevention ef-fect). We evaluated this hypothesis further by re-modelingwithout the morbidity variables. No change was seen inthe breastfeeding estimates, lending support to the hypoth-esis. All other factors remaining constant, infants who wereexclusively breastfed (EBF) in the ®rst 3 months werepredicted to be about 95 g heavier and 0.5 cm taller at 12months than those partially or not breastfed (based onmodels 2 and 4, Table 3). This is somewhat of an over-estimate because of the limitation of the mixed effectsmodels alluded to above.

The comparison between the EBF and PBF infants wasnot conclusive, because the latter group appears to gainweight and length faster than the EBF infants in the ®rst 6months of life (although this was not signi®cant), but hadslower weight gain than the EBF infants after 6 months(P< 0.01). We also showed an independent negative effectof complementary foods and drinks before 6 months onattained weight and length. This further strengthens theargument that complementary foods before 6 months of lifeare not necessary and are frequently detrimental. The lackof an effect of this variable after 6 months is dif®cult toexplain, but it is likely that this is a re¯ection of the limitedrange and overall inadequacy of complementary foodsoffered to infants after 6 months.

EBF, and possibly PBF, also appear to counter theoverall pattern of heavier babies growing even heavier inthis infant population. The weight growth rate of EBFbabies was not associated with starting body weight inboth halves of infancy (PBF babies behave similarly in thesecond half of infancy). In the younger ages we observedfaster gain in length among infants shorter at birth, but thiswas limited to EBF and PBF babies only. The most likelyexplanation of the observed effect on weight is that EBF(and PBF) allows even the lighter infants to grow at thesame rate as the heavier infants and that these infants areless restricted by the growth channel de®ned by their levelof intrauterine growth retardation. The effects on lengthreinforce this conclusion. What is especially remarkable,though, is how the effect of appropriate breastfeeding in the®rst 3 months of life was sustained throughout infancy. It isof course possible that appropriate behavior at one timeleads to better practices later and the observed sustainedeffect is a consequence of this association. This does notseem likely. In our study, infants exclusively breastfedin the ®rst 3 months did not receive greater number ofcomplementary foods from 6 months onwards. This argu-ment is further supported by the observed reduction in the®rst-6-months gains by EBF and PBF infants after 6 monthsof age. Studies in the Philippines, Sudan and Indonesiahave also demonstrated a positive effect of breastfeeding ongrowth beyond 6 months (Adair et al, 1993; Kusin et al,1991; Brush et al, 1993). In a study in Peru, infants whowere fully breastfed in the ®rst 3 months of life had thehighest cumulative weight gain by 12 months (Piwoz et al,1996).


175


We have observed a strong and negative associationbetween morbidity and growth, especially diarrhoea. Sinceboth were measured at the same time, it is dif®cult to makeconclusions regarding the direction of this association.However, the presence of an association between lengthand diarrhoea suggests that, in this instance, poor growthmay be a risk factor for morbidity. Previous studies haveprovided evidence of higher morbidity risk in the under-nourished (Zaman et al, 1996; Ballard and Neumann, 1995;Baqui et al, 1993a,b; El Samani et al, 1988; Bairagi et al,1987). Diarrhoea and respiratory infections have also beenshown to have a negative effect on growth (Adair et al,1993; Cole, 1989; Zumrawi et al, 1987; Black et al, 1984).Data collected at 3-month intervals, as is the case here, isnot conducive to an analysis of causality between nutri-tional status and morbidity. Furthermore, assessment ofdiarrhoea or ARI only at a few cross-sectional contacts willnecessarily lead to some misclassi®cation, which generallyreduces the ability to detect a real effect. Nevertheless, theinteresting element of the current analysis is the strengthand signi®cance of the association even after accounting forother variables, which is an indication of the strongrelationship between nutrition and morbidity in the livesof children in this population. It has been hypothesized thatincreased energy expenditure during illness, and reducedabsorption, withholding of food or poor appetite are impor-tant pathways through which illness in¯uences weight(Adair et al, 1993; Becker et al, 1991). Impaired bodydefenses to infection have been blamed for the higherillness experience of the undernourished (Chandra, 1991;Scrimshaw et al, 1968). Data from rural Bangladesh haveshown that poor nutritional status and cell-mediatedimmune de®ciency are independent risk factors for child-hood diarrhoea and respiratory infections (Zaman et al,1996; Baqui et al, 1993b).

Prevalence estimates of birth status may have beenaffected by the 181 newborns not enrolled in the studybut an effect on the observed associations is less likely,given the relatively small number of infants involved. Theinfants lost to follow-up (28%) were more likely to havemothers who were younger, primiparous, taller, poorer andborn outside Dhaka. Our analysis has show that, except fortaller mothers, all of these characteristics were negativelyassociated with infant growth. However, those whodropped out did not differ from those who completedfollow-up with respect to proportion LBW, SGA or preterm(data not presented). Since the losses to follow-up occurredat different ages and those infants contributed some datapoints prior to outmigration and because the particularanalysis method used allowed each of these infants tohave his=her own growth curve by borrowing informationfrom infants with shared characteristics, we do not expect amajor impact of the loss to follow-up on the study ®ndings,especially since we are interested in relationships withother variables rather than growth levels per se.

The availability of good quality birth weight data from arepresentative sample of births is a major advantage of ourstudy. Studies in developing country settings are usually

constrained to births in hospitals or in selected healthfacility catchment areas. Our study population is notserved by any particular health facility or organizationbut has access to a wide variety of services in urbanDhaka, including ICDDR,B, and nearly all of the birthsoccured at home.

In summary, this study highlights the importance of sizeand maturity at birth in determining infant growth in a poorurban community in Bangladesh and re-emphasizes theneed for improving fetal growth as a ®rst step towardsimproving childhood nutritional status. This argument ismade stronger by the observed pattern of bigger babiesgrowing even bigger. The data presented, including thegreater bene®t to smaller babies from breastfeeding, pro-vides renewed impetus to the efforts for the promotion ofbreastfeeding, especially exclusive breastfeeding in the ®rst6 months of life.

Acknowledgements ÐThis research was supported by the Royal Nether-

lands Government under the agreement no BD009602 with ICDDR,B, the

United States Agency for International Development (USAID) under

Cooperative Agreement no 399-0073-A-00-1054-02 with ICDDR,B and

under the Johns Hopkins Family Health and Child Survival Cooperative

Agreement (no HRN-A-00-96-9006-00), and the ICDDR,B: Centre for

Health and Population Research. The Centre is supported by over 30

countries and agencies which share its concern for the health and popula-

tion problems of developing countries.

References

Adair LS (1989): Low birth weight and intrauterine growth retardation inFilipino Infants. Pediatrics 84, 613 ± 622.

Adair L, Popkin BM, VanDerslice S, Akin J, Guilkey D, Black R, BriscoeJ, & Fleiger W (1993): Growth dynamics during the ®rst two years oflife: a prospective study in the Phillipines. Eur. J Clin. Nutr. 47, 42 ± 51.

Adair L & Popkin BM (1996): Low birth weight reduces the likelihood ofbreast-feeding among Filipino infants. J. Nutr. 126, 103 ± 112.

Alexander GR, Himes JH, Kaufman RB, Mor J & Kogan M (1996): AUnited States national reference for fetal growth. Obstet. Gynecol. 87,163 ± 168.

Arifeen SE (1997): Birth weight, intrauterine growth retardation andprematurity: a prospective study of infant growth and survival in theslums of Dhaka, Bangladesh. Doctor of Public Health dissertation.Baltimore, Maryland: Johns Hopkins University School of Hygiene andPublic Health.

Arifeen SE, Black RE, Caul®eld LE, Antelman G, Baqui AH, Nahar Q etal (2000): Infant growth patterns in the slums of Dhaka in relation tobirth weight, intrauterine growth retardation and prematurity, Am. J.Clin. Nutr. 72, 1010 ± 1017.

Bairagi R, Chowdhury MK, Kim YJ, Curlin GT, & Gray RH (1987): Theassociation between malnutrition and diarrhoea in rural Bangladesh.Int. J. Epidemiol. 16, 477 ± 481.

Ballard TJ & Neumann CG (1995): The effects of malnutrition, parentalliteracy and household crowding and acute lower respiratory infectionsin young Kenyan children. J. Trop. Pediatr. 41, 8 ± 13.

Baqui AH, Black RE, Sack RB, Chowdhury HR, Yunus M & Siddique AK(1993a): Malnutrition, cell-mediated immune de®ciency, and diarrhoea:a community-based longitudinal study in rural Bangladeshi children.Am. J. Epidemiol. 137, 355 ± 365.

Baqui AH, Sack RB, Black RE, Chowdhury HR, Yunus M & Siddique AK(1993b): Cell-mediated immune de®ciency and malnutrition are inde-pendent risk factors for persistent diarrhoea in Bangladeshi children.Am. J. Clin. Nutr. 58, 543 ± 548.


176


Baqui AH, Black RE, Arifeen, SE, Hill K, Mitra SN & Al Sabir A (1998):Causes of childhood deaths in Bangladesh: results of a nationwideverbal autopsy study. Bull. WHO. 76, 161 ± 171.

Barros FC, Victora CG, Vaughan JP, & Smith PG (1986): Birth weight andduration of breast-feeding: are the bene®cial effects of human milkbeing overestimated? Pediatrics 78, 656 ± 661.

Becker S, Black RE & Brown KH (1991): Relative effects of diarrhoea,fever, and dietary energy intake on weight gain in rural Bangladeshchildren. Am. J. Clin. Nutr. 53, 1499 ± 1503.

Binkin NJ, Yip R, Fleshood L & Trowbridge FL (1988): Birth weight andchildhood growth. Pediatrics 82, 828 ± 834.

Bittencourt SA, Leal MDC & Rivera J (1993): Diarrhoea and growthamong children under 18 months of age in Rio de Janeiro. Bull. PanAm. Health Org. 27, 135 ± 144.

Black RE, Brown KH & Becker S (1984): Effects of diarrhoea associatedwith speci®c enteropathogens on the growth of children in ruralBangladesh. Pediatrics 73, 799 ± 805.

Brush G, Harrison GA & Zumrawi FY (1993): A path analysis of somedeterminants of infant growth in Khartoum. Ann. Hum. Biol. 20, 381 ±387.

Capurro H, Konichezky S, Fonseca D & Caldeyro-Barcia R (1978): Asimpli®ed method for diagnosis of gestational age in the newborninfant. J. Pediatr. 93, 120 ± 122.

Casey PH, Kraemer HC, Bernbaum J, Tyson JE, Sells JC, Yogman MW &Bauer CR (1990): Growth patterns of low birth weight preterm infants:a longitudinal analysis of a large, varied sample. J. Pediatr. 117, 298 ±307.

Castillo-Duran C, Rodriguez A, Venegas G, Alvarez P & Icaza G (1995):Zinc supplementation and growth of infants born small for gestationalage. J. Pediatr. 127, 206 ± 211.

Caul®eld LE, Haas JD, Belizan JM, Rasmussen KM, & Edmonston B(1991): Differences in early postnatal morbidity risk by pattern of fetalgrowth in Argentina. Paed. Perinat. Epidemiol. 5, 263 ± 275.

Chandra RK (1991): 1990 McCollum Award lecture. Nutrition andimmunity: lessons from the past and insights into the future. Am. J.Clin. Nutr. 53, 1087 ± 1101.

Checkley W, Gilman RH, Epstein LD, Suarez M, Diaz JF, Cabrera L,Black RE & Sterling CR (1997): Asymptomatic and symptomaticcryptosporidiasis: their accute effect on weight gain in Peruvianchildren. Am. J. Epidemiol. 146, 156 ± 163.

Cole TJ (1989): Relating growth rate to environmental factors Ð metho-dological problems in the study of growth-infection interaction. ActaPaediatr. Scand. 350(Suppl), 14 ± 20.

Cruise MO (1973): A longitudinal study of the growth of low birth weightinfants. I. Velocity and distance growth, birth to 3 y. Pediatrics 51,620 ± 628.

Dewey KG (1998): Cross-cultural patterns of growth and nutritional statusof breast-fed infants. Am. J. Clin. Nutr. 67, 10 ± 17.

Diggle PJ, Liang K & Zeger SL (1994): Analysis of Longitudinal Data.Oxford: Clarendon Press.

El Samani EFZ, Willet WC, & Ware JH (1988): Association of malnutri-tion and diarrhoea in children aged under ®ve years. Am. J. Epidemiol.128, 93 ± 105.

Fawzi WW, Forman MR, Amalia L, Graubard BI, Naggan L & BerendesHW (1997): Maternal anthropometry and infant feeding practices inIsrael in relation to growth in infancy: the North African Infant FeedingStudy. Am. J. Clin. Nutr. 65, 1731 ± 1737.

Feacham RG & Koblinsky MA (1984): Interventions for the control ofdiarrhoeal diseases among young children: promotion of breast-feeding.Bull. WHO. 62, 271 ± 291.

Garn SM (1985): Relationship between birthweight and subsequent weightgain. Am. J. Clin. Nutr. 42, 57 ± 60.

Heikens GT, Scho®eld WN, Christie CDC, Gernay JG, Dawson S & TheKingston Project III (1993): The effects of high energy supplement andmetronidazole on malnourished children rehabilitated in the commu-nity: morbidity and growth. Eur. J. Clin. Nutr. 47, 174 ± 191.

Huttly SRA, Victora CG, Barros FC, Teixeira AMB & Vaughan JP (1990):The timing of nutritional status determination: implications for inter-ventions and growth monitoring. Eur. J. Clin. Nutr. 45, 85 ± 95.

Karniski W, Blair C & Vitucci JS (1987): The illusion of catch-up growthin premature infants. Use of the growth index and age correction. Am. J.Dis. Child. 141, 520 ± 526.

Kramer MS, Barr RG, Leduc DG, Boisjoly C & Pless IB (1985): Infantdeterminants of childhood weight and adiposity. J. Pediatr. 107, 104 ±107.

Kusin JA, Kardjati S, van Steenbergen WM & Renqvist UH (1991):Nutritional transition during infancy in East Java, Indonesia: 2. Alongitudinal study of growth in relation to the intake of breast milkand additional foods. Eur. J. Clin. Nutr. 45, 77 ± 84.

Lubchenco LO, Hansman C & Boyd E (1966): Intrauterine growth inlength and head circumference as estimated from live births at gesta-tional ages from 26 to 42 weeks. Pediatrics 37, 403 ± 408.

Lutter CK, Mora JO, Habicht JP, Rasmussen KM, Robson DS & HerreraMG (1990): Age-speci®c responsiveness of weight and length tonutritional supplementation. Am. J. Clin. Nutr. 51, 359 ± 364.

Mahmood DA & Feacham RG (1987): Feeding and nutritional statusamong infants in Basrah City, Iraq: a cross-sectional study. Hum. Nutr.Clin. Nutr. 41, 373 ± 381.

Marquis GS, Habicht JP, Lanata CF, Black RE, Rasmussen KM (1997):Breast milk or animal-product foods improve linear growth of Peruviantoddlers consuming marginal diets. Am. J. Clin. Nutr. 66, 1102 ± 1109.

Martines JC, Habicht JP, Ashworth A & Kirkwood BR (1994): Weaning inSouthern Brazil: is there a weanlings dilemma? J. Nutr. 124, 1189 ±1198.

Mata LJ, Urrutia JJ, Kronmal RA & Joplin C (1975): Survival and physicalgrowth in infancy and early childhood: study of birth weight andgestational age in a Guatemalan Indian village. Am. J. Dis. Child.129, 561 ± 566.

Peck RE, Marks JS, Dibley MJ, Lee S & Trowbridge FL (1987): Birthweight and subsequent growth among Navajo children. Public HealthRep. 102, 500 ± 507.

Persson LA (1985): Infant feeding and growth Ð a longitdinal study inthree Swedish communities. Ann. Human Biol. 12, 41 ± 52.

Piekkala P, Kero P, Sillanpaa & Erkkola R (1989): The somatic growth ofa regional birth cohort of 351 preterm infants during the ®rst two yearsof life. J. Perinat. Med. 17, 41 ± 49.

Piwoz EG, de Kanashiro HC, de Romana GL, Black RE & Brown KH(1996): Feeding practices and growth among low-income Peruvianinfants: a comparison of internationally-recommended de®nitions. Int.J. Epidemiol. 25, 103 ± 114.

Rosso P. (1989): Morbidity and mortality in intrauterine growth retarda-tion. In Intrauterine Growth Retardation, ed. J Senterre. Nestle Nutri-tion Workshop Series, Vol 18, pp 123 ± 140. New York: Raven Press.

Rowland MGM, Goh SGJ & Cole TJ (1988): Impact of infection on thegrowth of children from 0 to 2 y in a West African community. Am. J.Clin. Nutr. 47, 134 ± 138.

Schroeder DG, Martorell R, Rivera JA, Ruel MT & Habicht JP (1995):Age differences in the impact of nutritional supplementation on growth.J. Nutr. (Suppl 4), 1051S ± 1059S.

Scrimshaw NS, Taylor CE & Gordon JE (1968): Interactions of nutritionand infection. Geneva: World Health Organization.

Smith TA, Lehman D, Coakley C, Spooner V & Alpers MP (1991):Relationships between growth and acute lower-respiratory infections,in children aged < 6 y in a highland population of Papua New Guinea.Am. J. Clin. Nutr. 53, 963 ± 970.

Tenovuo A, Kero P, Piekkala P, Korvenranta H, Sillanpaa M & Erkkola R(1987): Growth of 519 small for gestational age infants during the ®rsttwo years of life. Acta Paediatr. Scand. 76, 636 ± 646

Urrusti J, Yoshida P, Velasco L, Frank S, Rosado A, Sosa A, Morales M,Yoshida T, & Metcoff J (1972): Human fetal growth retardation: I.Clinical features of sample with intrauterine growth retardation. Pedia-trics 50, 547 ± 558.

Victora CG, Barros FC, Vaughan JP, Martines JC & Beria JU (1987):Birthweight, socioeconomic status and growth of Brazilian infants. Ann.Hum. Biol. 14, 49 ± 57.

Victora CG, Morris SS, Barros FC, Horta BL, Weiderpass E & Tomasi E(1998): Breast-feeding and growth in Brazilian infants. Am. J. Clin.Nutr. 67, 452 ± 458.


177


Villar J & Belizan JM (1982): The timing factor in the pathophysiology ofintruterine growth retardation syndrome. Obstet. Gynecol. Surv 37,499 ± 506.

Villar J, Smeriglio V, Martorell R, Brown CH & Klein RE (1984):Heterogeneous growth and mental development of intrauterine growthretarded infants during the ®rst three years of life. Pediatrics 74, 783 ±791.

Villar J, Belizan J & Smeriglio V (1989): Postnatal experiences ofintruterine growth-retarded infants. In Intrauterine Growth Retardation,ed. J Senterre. Nestle Nutrition Workshop Series, Vol 18, pp 261 ± 277.New York: Raven Press.


178


Determinants of infant growth in the slums of Dhaka: size and maturity at birth, breastfeeding and morbidity

Documents