IIN PennyMEetal.lancet.vol365

7/31/2019 IIN PennyMEetal.lancet.vol365

1/10

Articles

IntroductionInadequate nutrition during the first 2 years of life canlead to morbidity in childhood, and is one of the mostimportant preventable risk factors for mortality.Undernutrition in childhood has been estimated to causehalf of all preventable deaths in infants worldwide.1 In

Peru, like in many developing and transitional societies,the most common nutritional problems are stunting2,3

and iron-deficiency anaemia.2,4,5 Inadequate nutritionmight be caused by lack of access to sufficient food or bylack of variety and quality, especially for micronutrientsof high bioavailability. However, caregivers might notmake the best use of available resources because ofcultural beliefs and practices, lack of knowledge ofthe best foods for young children even when availablein the home, and inappropriate advice.6,7 In thesecircumstances, interventions that provide additionalcomplementary food can prevent growth retardation,8

especially for children 612 months old.9

Large-scale educational interventions have also beeneffective in changing the way caregivers give food,

increasing dietary intake, and in improving childgrowth,8 although results have usually been based onpreintervention versus postintervention assessments oron comparisons with children not included in theeducation programme. Two controlled trials10,11 showedthat community-based culturally appropriate nutrition

education can improve infant-feeding practices, dietaryintake, and growth. However, interventions that areheavily dependent on community-based strategies arelimited in many countries because sustainabilitydepends on political expediency or on the continuedpresence of non-government organisations. In countrieswhere government health services provide wide coverageand are easily accessible, these services are a logical andmore sustainable channel for educational interventions.However, few randomised trials have measured theeffect of child-nutrition educational interventionsimplemented through health services. High turnover ofstaff and difficulty of training all health professionals

involved in counselling hinders interventions that focuson training individuals.

Lancet 2005; 365: 186372

Published online

May 10, 2005

DOI: 10.1016/S0140-6736(05)

66426-4

See Comment page 1832

Instituto de Investigacin

Nutricional, La Molina, Lima,

Peru (M E Penny MBChB,

H M Creed-Kanashiro MPhil,

M R Narro);Department ofInternational Health, Johns

Hopkins University, Bloomberg

School of Public Health,

Baltimore, MD, USA

(R C RobertPhD,

L E Caulfield PhD,

Prof R E Black MD)

Correspondence to:

Dr MaryPenny, Instituto de

Investigacin Nutricional,

Avenida La Molina 1885,

La Molina, Lima, Peru

[email protected]

www.thelancet.com Vol 365 May 28, 2005 1863

Effectiveness of an educational intervention delivered

through the health services to improve nutrition in youngchildren: a cluster-randomised controlled trial

Mary E Penny, Hilary M Creed-Kanashiro, Rebecca C Robert, M Rocio Narro, Laura E Caulfield, Robert E Black

SummaryBackground Malnutrition is the underlying cause of half of child mortality. Many programmes attempt to remedythis issue but there is a lack of evidence on effective ways to decrease child malnutrition.

MethodsWe did a cluster-randomised trial of an educational intervention in a poor periurban area (ie, shanty town)of Peru. Guided by formative research, the intervention aimed to enhance the quality and coverage of existingnutrition education and to introduce an accreditation system in six government health facilities compared with six

control facilities. The primary outcome measure was growth that was measured by weight, length, and Zscores forweight-for-age and length-for-age at age 18 months. Main secondary outcomes were the percentage of childrenreceiving recommended feeding practices and the 24-h dietary intake of energy, iron, and zinc from complementaryfood at ages 6, 9, 12, and 18 months. Analysis was by intention to treat.

FindingsWe enrolled a birth cohort of 187 infants from the catchment areas of intervention centres and 190 fromcontrol areas. Caregivers in intervention areas were more likely to report receiving nutrition advice from the healthservice than were caregivers in control health facilities (16 [52%] of 31 vs 9 [24%] of 37, p=002). At 6 months morebabies in intervention areas were fed nutrient-dense thick foods at lunch (a recommended complementary feedingpractice) than were controls (48 [31%] of 157 vs 29 [20%] of 147; difference between groups 19 [11%], p=003). Fewerchildren in intervention areas failed to meet dietary requirements for energy (8 months: 30 [18%] of 170 vs 45 [27%]of 167, p=004; 12 months: 64 [38%] of 168 vs 82 [49%] of 167, p=0043), iron (8 months: 155 [91%] of 170 vs 161[96%] of 167, 9 months: 152 [93%] of 163 vs 165 [99%] of 166, p=0047), and zinc (9 months: 125 [77%] of 163 vs 145

[87%] of 166, p=0012) than did controls. Children in control areas were more likely to have stunted growth (ie,length for age less than 2 SD below the reference population median) at 18 months than children in interventiongroups (26 [16%] of 165 vs 8 [5%] of 171; adjusted odds ratio 304 [95% CI 121764]). Adjusted mean changes inweight gain, length gain, and Zscores were all significantly better in the intervention area than in the control area.

Interpretation Improvement of nutrition education delivered through health services can decrease the prevalence ofstunted growth in childhood in areas where access to food is not a limiting factor.


2/10

Articles

To assess the potential benefit of health-service

educational interventions, trials that measure the effecton the general population are needed. We did a cluster-randomised effectiveness trial of a nutrition educationprogramme that focused on improving the service ratherthan improving individual service providers. The trial wasdone in poor periurban communities (shanty town areas)in Peru. We measured the effect of the 2-year study onfeeding practices, dietary intakes, and growth incatchment populations of 12 participating health services,and present results from a cohort of children born duringthe intervention and followed up until age 18 months.

MethodsParticipants

The study took place in Trujillo, a city located 400 kmnorth of Lima with a population12 of 600 000. Healthfacilities of the government of Peru serve populationsthat live in periurban shanty towns characterised byinhabitants with a low and insecure income, poorhousing, and by a general lack of one or more essentialservices (ie, piped water, reliable electricity supply, andsewage disposal). However, various nutritious foods areavailable to almost all families. Acute malnutrition (ie,low weight for length) is unusual in children in theseareas, but anaemia and growth faltering that leads tostunted growth are common. Patterns of growth inchildren are similar to those in periurban settlements in

Lima:

2,13,14

average birthweight

2

is 32 kg, 5% of childrenhave low birthweight, and weight and linear growth fallbehind international reference growth curves from age4 months. In children about 18 months or older, averagegrowth is similar to reference data. Although there canbe a catch-up period of weight gain, average lengthremains low for age.13

ProceduresThe intervention was done as a programme of theregional health authority, which appointed acommission to assist and monitor the intervention. Thedesign of the intervention, measurement of effect, andtechnical assistance during implementation were

provided by the project. We took care to ensure thatactivities of the intervention enhanced existing activitiesand were sustainable. This cluster-randomised controltrial was approved by the ethics committee of theInstituto de Investigacin Nutricional, Lima, Peru.Families who participated were informed of the studyprotocol and signed consent was obtained.

Health facilities serving periurban communities inTrujillo were eligible unless they were unique in someway (ie, had a characteristic not found in any othercentre). Facilities were divided into three types:community hospitals offering maternal and perinatalspecialist services; health centres with medical staffalways in attendance; and health centres with morelimited services. Facilities were paired for similar

socioeconomic conditions in catchment populations (on

the basis of local knowledge of personnel in the healthregion).Randomisation was done before formative research to

avoid it acting as an intervention. Starting with thecommunity hospitals, one of each pair (ie, two centres ofthe same type and broadly similar characteristics) wasassigned to intervention or control by the investigator(MP) tossing a coin in the presence of the local healthauthorities. Because families might attend a close-byhealth facility rather than the intended catchment facility,we excluded a pair if the randomisation resulted in acontrol site being directly adjacent to an intervention site.Randomisation of pairs continued until we hadcompleted the predetermined number of six pairs of

facilities. The seventh pair was held in reserve.We studied the interventions implementation process

and the effect of the intervention on child outcomes. Theassessment of the processwhich involved use ofstructured observations in the health facilities,interviews of caregivers, and a cross-sectional surveywill be published elsewhere. To assess the effect of theintervention, a cohort of children was followed up frombirth to age 18 months, and a final survey was done atthe end of the experimental phase. Data were collectedby project field workers who were not involved in thedelivery of the intervention.

The intervention aimed to raise the profile of nutrition

in the health facilities and to integrate nutrition servicesinto existing child-oriented national programmes such asimmunisation, monitoring of growth and development,and management of acute respiratory infections anddiarrhoea. We aimed to enhance the quality of nutritioncounselling through training and provision of simple,standardised, age-appropriate messages to be used at allpoints of contact with young children in the facility.Materials available in health facilities were adapted forthe study and provided as flip charts and single-pagerecipe fliers. Three key messages were designed anddisseminated among all staff in the facilities that had anycontact with caregivers of young children. Thesemessages were: a thick puree satisfies and nourishes your

baby, equivalent to three portions of soup: at each mealgive puree or thick-food preparation first; add a specialfood to your babys serving: (chicken) liver, egg, or fish;and teach your child to eat with love, patience, and goodhumour. Facilities were assisted in developing their ownprotocols for use of the educational materials, and clinicalhistory forms were designed to help prompt physicians toinclude brief questions and advise on nutrition. Theintervention included demonstrations of the preparationof complementary foods, and group sessions forcaregivers of children of similar ages were added toenhance the coverage and the nutrition content of thegrowth-and-development-monitoring programme in well-baby clinics. The intervention also provided training toimprove anthropometry skills in health-care workers.

1864 www.thelancet.com Vol 365 May 28, 2005


3/10

Articles

We introduced an accreditation scheme as a

mechanism for institutional change, two rounds ofwhich were done during the course of the intervention.Accreditation was done by two local health professionals(who were selected by the health region and the projectteam) and by project workers (who were trained by theproject team). Accreditation was based on the satisfactionof previously defined criteria that measured the healthservice compliance with the intervention and was doneby a review of health-facility records, observation ofcontact with patients, interviews with caregivers of youngchildren on leaving the facility, and by a few home visitsto caregivers who had visited the facility in the preceding2 weeks. Structured instruments with a defined formatwere used throughout, and results were given to the

facilities immediately after the accreditation visit.Accredited facilities received public recognition and acommemorative plaque to place in the clinic.

We hypothesised that the intervention would lead toimproved feeding practices, dietary intakes, and growth ofchildren in the catchment areas of the intervention healthfacilities. The primary outcome was growth measured byweight, length, and Z scores for weight-for-age andlength-for-age at age 18 months. Secondary outcomeswere the proportion of children receiving recommendedfeeding practices and the 24-h dietary intake of energy,iron, and zinc from complementary foods at ages 6, 9, 12,and 18 months. We enrolled a cohort of newborns after a

census of the catchment population of each health facility.In small catchment areas, the whole area was visited toidentify all pregnant women; larger areas were dividedinto geographical districts defined by the health facilitiesand a subsample was enrolled from each district. In everydistrict, a block was chosen at random and then a singlehouse was randomly selected as a starting point for thecensus. When babies were born in these households theirparents were invited to participate. Additional babies wereidentified by extending the census to contiguous areas.

The first training activities of the intervention started inMay, 1999. Newborns who were found at home, whowere aged 10 days or younger, who had no knowncongenital malformation or chronic condition that could

affect growth, and whose parents gave written informedconsent were enrolled between Aug 13, 1999, and Feb 10,2000, and followed up until 18 months of age. Familieswere not told whether they were in the intervention orcontrol group.

Fieldworkers visited the families as soon as possibleafter birth and when the baby was 3, 4, 6, 8, 9, 12, 15, and18 months oldcrucial ages at which to assess nutrition,feeding practices, and growth. At the first visit,fieldworkers collected information on familycomposition and socioeconomic conditions and, at eachsubsequent visit, questioned families about the use ofhealth services and exposure to nutrition messagesduring the period between interviews; they also used astandardised, structured-observation technique to assess

home hygiene. Reported signs of illness during the

previous 24 h were used as an estimate of overallpopulation morbidity.At each visit, children were weighed on clock balances

accurate to 100 g that were calibrated weekly, andrecumbent length was measured with locally made, rigidstadiometers accurate to 01 cm. Data collectors weretrained and standardised according to WHO guidelines15

before the study started. Questions about age-specificfeeding practices were asked at each visit. At 6, 8, 9, 12,15, and 18 months, trained nutritionists did a secondinterview to estimate dietary intakes using quantitative24-h recall. Weighing of food portions with digitalbalances (maximum weight 2 kg and precision 1 g), useof household utensils, and locally made food models

(representing different sizes of food or of food portions)helped estimate portion sizes and the ingredients ofmixed dishes consumed by the child.

Statistical analysisFinancial and practical factors limited the number ofhealth services in which we could work, and six healthfacilities were thus thought to be the minimumacceptable number of health services per group. We didnot have information from the population to calculateintracluster correlation, but did expect the design effect tobe quite low. The sample size for the number ofindividuals in the cohort was calculated for 80% power

and 95% CI for the primary outcomes. Based on thefrequency of reported exclusive breastfeeding in coastalcities excluding Lima,2 we anticipated differences of 15%in exclusive breastfeeding (increase from 15% to 30% at 2months [n=133 per randomised group], and from 10% to25% at 4 months [112 per group]). Furthermore, weexpected a 15% difference in the frequency ofrecommended complementary feeding (ie, breastfeedingplus two or more energy-dense preparations, from 15%to 30% [133 per group]), and a minimum difference of112 kcals per day, 2 mg iron per day, and 1 mg zinc perday in children age 612 months. These estimates werebased on observed feeding practices in similar periurbancommunities in Lima.16 A difference of 05 cm at 1 year

was predicted for linear growth.13To account for the cluster design and the anticipated

(low) loss to follow-up, we increased the sample size by25% and aimed to enrol 180 children per group (30 percluster). We did not make an allowance for the differentsizes of cluster catchment areas.

Data were collected on forms that were previouslypiloted, checked by field supervisors, entered in purpose-designed programmes (which included range andconsistency checks), and analysed with SPSS version 10.Socioeconomic information was analysed by use ofprincipal-component analysis, and three clusters ofassociated variables were identified: housing,possessions, and educational level of the parents. Ahygiene score was calculated from the observations of the



4/10

Articles

household. Food intake was converted to nutrients by useof Peruvian food-composition tables,17 which include datafor food composition from other Latin American and UStables where necessary. We also compared data for dietary

intake with recommended daily intakes forcomplementary foods.6,18,19 Anthropometric data wereconverted to Zscores by use of reference values.20

Analysis was by intention to treat, and children whomoved away were analysed as part of the community inwhich they were enrolled unless they were lost to follow-up because they moved out of the study area completely.Outcomes reported here all pertain to the individuallevel rather than at the cluster level. Outcomes werepredefined as: knowledge of key feeding practices andmessages; actual desirable feeding practices (defined asan age-dependent compositeeg, exclusive breast-feeding for children younger than 6 months, breast milkplus two thick foods and one animal product per day forchildren older than 6 months); and dietary intakes from

complementary foods for energy, macronutrients,

micronutrients, and foods from an animal source forchildren older than 6 months. For morbidity outcomes:diarrhoea was defined as three or more liquid or semi-liquid stools in 24 h; fever as reported by parents; andanorexia as a reduction in appetite that was a changefrom the childs usual state. Childrens visits to thehealth facilities were classified as mainly illness-relatedor as preventive (ie, attendance at well-baby clinic,vaccination, or nutrition consultation). Assessment ofgrowth was by comparison of final attained weight,height, and associated Zscores at 18 months of age.

Results were compared by use of ANOVA withtransformation if needed; when not possible, non-parametric statistical tests were applied. Significance was

accepted at p005. Differences between groups wereassessed after adjustment for biological (ie, birthweightand sex) and socioeconomic covariates. Random-effectsmodels were applied to account for cluster randomisation.We did a permutation test to test the likelihood of type Ierror (ie, the chance of an equal or greater difference inoutcome if the data had been generated by a differentrandomisation process)21 by reanalysing of the outcomesof all possible randomisations by reassignment of healthcentres to intervention and control.

Role of the funding sourceThe sponsors of the study had no role in the study design,data analysis, data interpretation, or writing of the report.The corresponding author had full access to all the data


Intervention Control

(n=187) (n=190)

Number of males 87 (54%) 98 (48%)

Housing score* 342 (105) 293 (135)

Mothers education level

(number completing second ary school) 97 (52%) 6 8 (36 %)

Fathers education level (number 109 of 167 (65%) 103 of 163 (63%)

completing secondary school)

Possessions 275 (113) 234 (125)

Household spending on food/person/day 297 (151) 270 (121)

Crowding 072 (028) 064 (028)

Hygiene score 252 (074) 193 (091)

Birth order of child:

First 94 (50%) 83 (44%)

Second 39 (21%) 49 (26%)

Third or over 54 (29%) 58 (31%)

Mean age (days) 015 (006) 015 (005)

Birthweight (kg) 341 (045) 335 (046)

Weight (kg) 324 (045) 322 (043)Length (cm) 496 (191) 498 (219)

Number participating in community 59 (32%) 62 (33%)

organisation||

Data are number of people (%) or mean (SD). *Source of drinking water, storage of

drinking water, sewage disposal, floor, and wall material of house. TV, radio,

refrigerator, and food blender. Amount spent on food in soles (local currency) per

person per day. Number of residents in house/number of rooms excluding bathroom.

Based on observations of the cleanliness of the home, specifically the state of the

floor and whether human faeces were visible around the house. ||For example, a

community kitchen or mothersclub.

Table 1: Characteristics of children and their household at enrolment

21 health facilities

assessed for eligibility

16 paired

5 excluded

before pairing

14 randomised

2 excluded

(1 pair)

12 included in study

6 assigned education

intervention

357 pregnant women

identified

2 excluded

(1 pair)

170 excluded

187 babies enrolled

187 babies analysed

1 died

3 withdrew

consent

12 moved away

6 assigned control

434 pregnant women

identified

244 excluded

190 babies enrolled

190 babies analysed

1 71 c omple ted fo ll ow -u p 1 67 co mplet ed fo ll ow -u p

3 died

5 withdrew

consent

13 moved away2 unknown

Figure 1: Trial profile


5/10

Articles

and full responsibility for the decision to submit for

publication.

Results21 health facilities were identified: three communityhospitals, one of which was excluded because it had justparticipated in a health-services intervention with anutrition component; five health centres, one of whichwas excluded because it shared administration with areligious organisation; and 13 health posts, two of whichwere excluded because they were thought too dangerousfor fieldwork and one because it was established for adisplaced population and was therefore regarded asunique. Two were excluded after pairing because therequired number had been reached and two more were

excluded because they were directly adjacent to eachother (figure 1). There was one community hospital, twohealth centres, and three health posts in each studygroup.

791 pregnant women were identified by house-to-house census in the catchment areas, 377 (48%) ofwhom were enrolled (figure 1). The main reasons for notbeing enrolled were that the needed sample size hadbeen achieved or that the baby had been born beforepredicted and was outside the age criterion; 12 mothersrefused to participate and two had babies with congenitalmalformations. During the 18 months, 75 families leftthe study for varying lengths of time because they moved

out of the area or decided not to continue. Some of thesefamilies later rejoined the study and at 18 months all but39 children and their caregivers were available toparticipate (figure 1). All clusters were included in theanalysis. The health facilities had been paired withrespect to the services offered. Table 1 shows thecharacteristics of children and families at enrolment

Children in the intervention group made significantlymore preventive health-care visits than did those in thecontrol group (973 visits vs 817 visits, p00001) over18 months. This finding was significant in the monthspreceding the visits when children were aged 3, 6, 12, 15,and 18 months. The number of visits related to illnessand the frequency of signs of morbidity 24 h before every

interview did not differ between intervention and controlgroups (data not shown).

Twice as many mothers in the intervention groupreported receiving nutritional advice shortly after birththan did those in the control group (16 [52%] of 31 vs 9[24%] of 37, p=002). These numbers increased in bothstudy groups with the age of the child, but reports weresignificantly greater in the intervention group than incontrols only at 4 months and 18 months (p=0002).Twice as many mothers in the intervention group as inthe control group who reported taking their child to ahealth centre had stated that they had receivednutritional advice.

The results with respect to caregivers knowledge ofage-specific feeding practices are shown in table 2. In


Intervention Control p (unadjusted) p (adjusted)

What foods or liquids does a baby need for the first months?Birth 138 of 187 (74%) 140 of 190 (74%) 0980

4 months 141 of 171 (825%) 130 of 170 (77%) 0171

How can a mother have more breastmilk?

Birth 16 of 187 (9%) 14 of 190 (7%) 0670 No significant

4 months 27 of 171 (16%) 11 of 170 (7%) 0006 covariables

What do you do if the child has colic or gases?

Birth 17 of 187 (9%) 17 of 190 (9%) 0961 No significant


Name three foods that you think are the best for an infant 78 months old

8 months 108 of 170 (64%) 81 of 169 (48%) 0004 0011*

12 months 113 of 170 (67%) 78 of 169 (46%) 000016 No significant


18 months 112 of 171 (66%) 91 of 167 (55%) 0039 0107

When you give the main meal to a child, what is the food you give first?



15 months 84 of 167 (50%) 56 of 169 (33%) 000118 months 87 of 171 (51%) 54 of 167 (32%) 0001 001

No significant covariables

What would you do if a child does not want to eat?

8 months 66 of 170 (39%) 33 of 169 (20%) 00001 00001


15 months 73 of 167 (43%) 49 of 169 (29%) 0005 Covariables

18 months 75 of 171 (44%) 45 of 167 (27%) 0001 0.006

Data are number (%). *=Adjusted for birthweight. =adjusting for housing. =adjusted for sex. =adjusted for parents education.

Table 2: Number of carers who correctly answered questions on age-specific feeding practices

Intervention Control p (unadjusted) p (adjusted)

Currently breast feeding*

3 months 171 of 174 (99%) 172 of 173 (99%) 0317

6 m ont hs 1 63 o f 1 72 ( 95 %) 1 57 o f 1 65 ( 95 %) 0 8 72

8 m ont hs 1 57 o f 1 69 ( 93 %) 1 58 o f 1 68 ( 94 %) 0 6 70

9 m ont hs 1 54 o f 1 64 ( 94 %) 1 50 o f 1 65 ( 91 %) 0 3 06

12 months 144 of 159 (91%) 150 of 162 (93%) 0514

15 months 117 of 157 (75%) 131 of 155 (85%) 0029

18 months 100 of 145 (69%) 113 of 150 (75%) 0223 No significant covariables

Exclusively breastfeeding*

Birth 143 of 187 (77%) 144 of 190 (76%) 0877

3 months 94 of 174 (54%) 105 of 174 (60%) 0233

4 months 76 of 171 (44%) 81 of 170 (48%) 0553

Eating thick food first at main meal


8 months 44 of 166 (21%) 34 of 163 (27%) 0229


12 months 69 of 166 (42%) 43 of 165 (26%) 0003

15 months 70 of 166 (42%) 56 of 166 (34%) 0114

18 months 75 of 171 (44%) 63 of 167 (38%) 0252

Consuming egg, chicken liver or fish

6 months 111 of 171 (65%) 84 of 165 (51%) 0009 0009

8 months 104 of 170 (61%) 81 of 167 (49%) 0019 0027

9 months 93 of 163 (57%) 80 of 166 (48%) 0108

12 months 94 of 168 (56%) 90 of 167 (54%) 0705

15 months 84 of 168 (50%) 88 of 167 (53%) 0622

18 months 109 of 171 (64%) 96 of 168 (57%) 0215

Breast milktwo thick preparationsanimal source food


8 months 40 of 170 (24%) 31 of 167 (19%) 0264

9 months 55 of 163 (34%) 38 of 166 (23%) 0029

12 months 61 of 168 (36%) 59 of 167 (35%) 0852

15 months 60 of 168 (36%) 68 of 167 (41%) 0347

18 months 72 of 171 (42%) 71 of 168 (42%) 0977

*Questionnaire data. 24-h dietary recall data. Adjusted for mothers age. Adjusted for birthweight.

Table 3: Feeding practices (24-h recall)


6/10

Articles

both groups, mothers knew the importance of

breastfeeding from birth at time of delivery. At4 months, when asked what could be done to producemore breastmilk, significantly more mothers in theintervention group answered more frequent suckling. Atthe 4-month visit, more women in the interventiongroup than the control group reported that if her babyhad colic, she could drink the digestive teas rather thangiving them directly to her baby (table 2). Moreover,more mothers in the intervention group respondedcorrectly to questions asking them to name threeimportant foods (ie, chicken liver, eggs, and fish) for achild aged 78 months than did mothers in the controlgroup. More caregivers in the intervention groupresponded that they would give thick foods first at main

meals than did those in the control group. Morecaregivers in the intervention group said they wouldencourage their child to eat if they had a reducedappetite than did those in the control group (table 2).

Nearly all women reported breastfeeding their baby,and there were no significant differences betweengroups (table 3). Using the WHO definition of theprevious 24 h, 760% of infants in each study groupwere exclusively breastfeeding at enrolment. Thefrequency of exclusive breastfeeding dropped to 540%(intervention) and 603% (control) at 3 months, and to444% and 476%, respectively, at 4 months (table 3).The reported number of breastfeeds in 24 h only differedbetween groups at 15 months, when intervention babieswere breastfed slightly more often than were controls

(117 feeds vs 100 feeds, p=0036 after log

transformation).Complementary feeding practices differed betweengroups. One key message of the intervention was to offera thick food (eg, pure or main course) to a child at themain meal before the commonly offered soup (whichhas low energy and nutrient densities compared withthick food). From the 24-h dietary recall, consumption ofthick food first was significantly greater in theintervention group than in the control group at6 months, 9 months, and 12 months. Another importantmessage of the intervention was to add a source ofanimal protein (ie, egg, chicken liver, or fish) to thechilds meals everyday. A significantly higher proportionof children in the intervention group received chicken

liver, fish, or egg than did controls at age 6 months and 8months. Enhanced intake of animal protein was causedmainly by increased frequency of chicken-liverconsumption. The message to motivate caregivers toteach their children to eat with love, patience, and goodhumour could not be assessed by interview.

One of the main study objectives was to increase thenumber of children older than 6 months who receivedbreast milk plus two thick preparations and one foodproduct from an animal source. However, the differencebetween groups was significant only at 9 months (table 3).

Mean energy intake from complementary foods wassignificantly higher in the intervention group than in the

control group at 18 months (figure 2). The differencesremained significant after adjustment for householdsocioeconomic characteristics. For all ages, mean energyintake in intervention and control groups exceeded therecommended intakes for complementary foods,assuming average breastmilk intake. However, morechildren in the control group did not meet 80% ofrecommended energy intake per day compared withthose in the intervention group at 8 months (45 [27%] of167 vs 30 [18%] of 170, p=0040), 12 months (82 [49%] of167 vs 64 [38%] of 168, p=0043), and at 15 months (51[31%] of 167 vs 35 [21%] of 168, p=0042). Adjustmentfor baseline characteristics eliminated the significance ofthe difference at 8 months (p=0176), 12 months

(p=0176 after adjustment for housing), and at15 months (p=0090 after adjusting for parental age).

Energy intake from animal sources was used as anindicator of food intake from an animal source anddietary quality. Children assigned to the interventiongroup had significantly higher intakes of energy fromanimal sources than did those assigned to the controlgroups at age 15 months and 18 months (figure 2).Adjustment for household characteristics in themultivariate model reduced significance at 15 months top=0082, but the difference in the adjusted meansremained significant at 18 months (p=0001). Theincrease in energy from animal sources was the result ofan increase in the use of animal-source foods rather thanan increase in the overall volume of food given.


Intervention

Control

Dietary energy from complementary foods

0

100

200

300

400

500

600

700

800

900

1000

0

50

100

150

200

250

300

350

400

Dietary iron intake from complementary foods

0

1

2

3

4

5

6

7

6 8 9 12 15 18 6 8 9 12 15 18

Dietary zinc intake from complementary foods

0

1

2

3

4

5

Dietary energy from animal source

complementary foods

Age of children

(completed months)

Age of children

(completed months)

Energyintake(kcal/24h)

Iron(kcal/24h)

Energyintake(kcal/24h)

Iron(kcal/24h)

Figure 2: Unadjusted mean energy and nutrient intakes from complementary foods (24-h recall)


7/10

Articles

Mean total iron intake from complementary food was

well below recommended intakes for every age group,but was higher in the intervention than in the controlgroup. Unadjusted differences were significant at 8, 9,and 18 months (figure 2), but the significance was nolonger apparent after adjustment for householdsocioeconomic variables (p=031 at 8 months, 015 at9 months, and 010 at 18 months). More than 90% of allchildren younger than 12 months did not have 80% ofdaily-recommended intakes for iron, diminishing toabout 40% of children at 18 months old. Differencesbetween intervention group and control group weresignificant at 8 months (155 [91%] of 170 vs 161 [96%] of168, p=0047) and at 9 months (152 [93%] of 163 vs 165[99%] of 167, p=0003 and at 18 months (63 [37%] of 171

vs 80 [48%] of 168, p=0045).Iron intake from good sources of haem (ie, meat,

poultry, and fish), although low, were non-significantlyhigher in the intervention group than in controls at6 months (mean 075 mg per day [SD 088] vs 064 mgper day [097]; p=0275 [ANOVA]) and at 8 months(082 [136] mg per day vs 066 mg per day [118];p=0257 [ANOVA]), when adjusted for socioeconomicvariables. However, unadjusted values by use of Mann-Whitney test showed significant differences. Thenutrient densities for iron and zinc were consistentlylower than recommended values6 and there were nodifferences in nutrient density for iron or zinc at any age

(data not shown).Mean zinc intake was also lower in participants thanthat recommended, but was significantly higher in theintervention group at 9 months than in the controlgroup (233 mg per day [SD 189] vs 185 mg per day[160], p=0014, figure 2). This difference remainedsignificant when mean intakes were adjusted forparents education levels in the model (p=0035). Theproportion of children not meeting 80% ofrecommended intakes was significantly lower in theintervention group than in the control group at9 months (125 [77%] of 163 vs 145 [87%] of 167,p=0012). For other micronutrients, calcium intakefrom complementary foods was also significantly higher

at 18 months in the intervention group than in controls(adjusted mean 526 mg per day vs 393 mg per day,p=0007 after adjustment for parental age).

Growth was similar to that described for the Peruvianpopulation2,13ie, parallel to or above the reference18 forthe first 46 months followed by a reduction in meanlength and weight increments. Figure 3 showsunadjusted, main-effect growth outcomes over18 months. Linear growth was much the same in bothintervention and control children for the first 6 months,followed by a divergence between groups and lessstunting in the intervention group. The interventionreduced stunting in linear growth (figure 4 and table 4).Linear-growth velocity, measured as the group meangradient of the regression line of the Zscores for length-

for-ageage for each individual, was significantlygreater in the intervention group than in controls, evenafter adjustment for socioeconomic status, hygienescore, and birthweight variables and after application ofthe random-effects model in recognition of the clusterdesign.

At 18 months, children in the intervention group were1 cm taller (p=00003) and three-times less likely(p=0018) to be stunted than were children in the control


Figure 3: Changes in weight-for-age and length-for-ageZ scores in children in

the cohort

Error bars=SE.

Age (months)

0

10

5

15

20

Proportionofchildren(%)

0 2 4 6 8 15 18

Intervention

Control

Figure 4: Cumulative rate of stunting from 0 to 18 months

Age (months)

Weight for age

15

05

10

0

05

10

15

0 3 6 9 12 15 18 21

Zs

core

Length for age

15

05

10

0

05

10

15

Zs

core

Intervention

Control


8/10

Articles

group, even after correcting for the effect of birthweight(figure 4). Mean weight gains also differed betweengroups: children in the intervention group at 18 months

weighed 295 g more than did controls (p=0014) beforeadjustment and 199 g more (p=0093) after adjustmentfor birthweight and socioeconomic factors.

DiscussionWe have shown improved knowledge, preventive health-care-seeking behaviour, feeding practices, dietary intake,ponderal growth, and linear growth in a cohort ofchildren living in the catchment area of health facilitiesthat participated in an educational intervention toimprove nutrition in young children. The interventionreduced the rate of stunting by more than two thirds.These results add to evidence that nutrition education

without the provision of food supplements can improvethe dietary intake of young children and improve growth.Several randomised control trials of interventions that

have included nutrition education combined with otherstrategies found improvements in growth and preventionof malnutrition, but the effect of the nutrition-educationcomponent cannot be distinguished.2224 A non-randomised community-based education programme inChina10 increased the frequency of consumption of foodsfrom an animal source and improved infant growth. Acluster-randomised trial in India11 reported significantlyincreased exclusive-breastfeeding rates, a low frequencyof diarrhoea, and a growth benefit in boys but not girlscompared with a control group.25 The Indian study had a

similar design to our study, in that the intervention wasdeveloped after formative research and implementedthrough existing services. The positive effects recorded intwo different circumstances suggest that thisintervention might be an effective programme ofnutrition education, which could have positive benefitsthat are measurable in the population served by thetargeted health programme.

The strengths of this study are the randomised design,comparison of six intervention health facilities with sixcontrol facilities, and that the number of childrenenrolled exceeded that estimated to ensure sufficientpower to test the hypotheses. Refusal to participate andloss to follow-up were low. Randomisation ensured thatcontiguous communities were not assigned to different

study groups, thus reducing intervention contamination.Moreover, the assessment team was completelyindependent of the intervention. The intervention

communities had a further round of surveys in themidterm as part of the process assessment, but care wastaken to avoid giving any nutrition or health adviceduring the interviews and only nine children in themidterm survey were also part of the cohort.

There were some inevitable weaknesses in the studydesign. The study could not be blinded, which couldhave led to bias. However, data collection wasstandardised, interviews were structured, andinterviewers rotated between intervention and controlareas to limit any bias that might result from the sameteam always interviewing intervention or controlfamilies. Nevertheless, knowledge of the group could

have influenced data collectors interpretation ofresponses or the recording of dietary-recall data, but thisknowledge is unlikely to have affected weight or heightmeasurements.

A further limitation was the socioeconomic differencesbetween families in the intervention and control groups.We had found that some socioeconomic characteristicshousing score, educational level of parents, hygiene score,and birthweightaffected variation in growth outcomesin the population. This information allowed us to developa predefined analysis plan using the appropriate variablesin the multivariate analysis of differences in the growth ofthe cohort. In the final model, differences in growthbetween the groups remained significant after

adjustment. After adjustment in the multivariate analysis,differences between groups for dietary intake remainedsignificant, providing clear evidence of the effect of theintervention on dietary intakes.

The magnitude of the difference in growth betweenthe groups is greater than might be expected from thesignificant, but modest, differences in feeding practicesand dietary intakes. Emphasis on preventive nutrition(which led to differences in the groups at an early agebefore adverse nutritional effects occurred) might havebeen crucial, or a small but sustained difference innutrition throughout early life could have resulted infairly large effects in the prevention of growthretardation. Errors in the estimation of outcomes suchas behaviours and dietary intake make the detection of


Intervention Control Undadjusted difference p Adjusted difference p

Mean weight at 18 months (kg) 1077 (116) 1048 (102 0295 (0061 to 0529) 0014 0199 (0033 to 0431) 0093Mean length at 18 months (cm) 7936 (274) 7829 (266) 1068 (0488 to 1648) 0 0 00 3 0 7 14 ( 0 14 6 t o 1 2 82 ) 0 0 14

Mean z-score weight for age at 18 months 033 (090) 062 (083) 0285 (0099 to 0471) 0003 0194 (0008 to 038) 0041

Mean z-score length for age at 18 months 081 (080) 119 (083) 0386 (0209 to 0562) 0 0 00 1 0 2 72 ( 0 09 9 t o 0 4 45 ) 0 0 02

Mean z-score weight for length at 18 months 015 (087) 005 (079) 0091 (0089 to 0271) 0319 0048 (0139 to 0237) 0609

L in ear g ro wt h ve lo cit y* 0 0 31 ( 00 43 ) 0 05 7 ( 0 05 ) 0 0 26 ( 0 01 6 t o 0 0 36 ) 00001 0021 (0012 to 0031) 00001

Number of children with stunting at 18 months 8 of 171 (47%) 26 of 165 (158%) 3811 (1672 to 8689) 0001 3035 (1207 to 7636) 0018

Data are mean (SD) or mean (95% CI). *Mean of the regression gradient for length for age vs age. Odds ratio (95% CI).

Table 4: Child growth at 18 months by intervention group


9/10

Articles

intervention effects more problematic than for

outcomes such as nutritional status or growth.Issues of sustainability and generalisability need to beaddressed in all public-health interventions. Our studytook place in a periurban population in a coastalPeruvian city. The population has ready access to agenerally well-organised and professional healthservicea likely prerequisite for an educationalintervention. Furthermore, food availability alone is nota limiting factor because almost all families are able tomeet recommended intakes for dietary energy.15

However, cultural preferences favour foods with a low-energy density such as soups and broths,6 which reducethe total energy intake of infants.16 Foods from animalsources that are high in bioavailable micronutrients are

fairly expensive, but sources such as those selected forpromotion in this intervention (ie, chicken liver, eggs,and fish) are affordable. In some areas, including thePeruvian highlands, food availability may be limited andeducation alone might have no effect if notcomplemented by food assistance. However, even iflimited to similar periurban populations described here,this intervention could benefit large populations becausethe greatest number of poor people live in these urbanareas in Peru.12 Similar situations are found in othercities in Latin America and other developing countries.

Our study selected government health services as themost widely used and sustainable vehicle for nutrition

education because even in a context of changeablegovernment policies and international donor priorities,local government health services continue to providehealth care (including immunisations and monitoringof growth and development) to poor areas. Despite theimportance and potential of interventions through thehealth services, few randomised controlled trials ofeducational interventions in developing countries havebeen done. The Baby Friendly Hospital Initiativeincreased exclusive breastfeeding in Belarus,24 and theIntegrated Management of Childhood Illnesses (IMCI)26

is another approach that uses health services to improvenutrition. A randomised trial27 of the nutritioncomponent of IMCI in Brazil showed limited benefits:

the intervention improved weight gain in children olderthan 12 months whose caregivers had been counselledby trained doctors, but there was no effect on the lineargrowth of children younger than 12 months. Becauseassessment was limited to the children of caregiverswho had received counselling from trained doctors, theextent of benefit in the overall population is unknown. 27

The intervention in the Brazilian study focused ontraining medical staff; by contrast, our interventionlooked at the service offered by the health facility overalland the effect was measured in the whole infantpopulation served by the facility.

We were careful to limit the intervention tosustainable enhancement of existing services. No newpersonnel were added, and material benefits given to

the centres were kept to a minimum, mainly in the

form of educational materials. The programme aimedto improve services rather than train individuals whomay move on and be replaced. However, participationof the health facility in a project probably generatedsome enthusiasm, and even if the project did notdepend on investing in individuals, the role of highlymotivated individuals is difficult to avoid.

Malnutrition is a major cause of child morbidity andmortality, and effective interventions are urgentlyneeded to prevent growth faltering in young children.Our educational intervention prevented stunting, aform of chronic malnutrition that occurs in more than15% of infants in this population. Research is needed todetermine the sustainability of the intervention in

Trujillo, and the generalisability of the interventionstrategy to similar settings in Peru and elsewhere.

Contributors

M E Penny and H M Creed-Kanashiro conceived and participated in theproject, interpretated the data, and the wrote the article. L E Caulfieldand R E Black advised on study design and implementation, andparticipated in the interpretation of the data and the writing of thearticle. M R Narro and R C Robert participated in the design of theintervention, assessment, and revision of the article.

Conflict of interest statement

We declare that we have no conflict of interest.

Acknowledgments

This project was supported by the Family Health and Child SurvivalCooperative Agreement between the United States Agency forInternational Development and Department of International Health,

Johns Hopkins Bloomberg School of Public Health, MD, USA.Thanks to Victor Montoya, Claudia Cuba, and Margot Marin for datamanagement and statistical analysis. The dedicated efforts of the fieldstaff and the secretarial support of the Instituto de InvestigacinNutricional, Peru, are much appreciated. The study was made possibleby the personnel of the participating health services, the local healthauthorities (especially Maria Gallo and Angela Cabanillas), and thecommission established by the regional health authority(Manuel Burgos, Sonia Castillo, Gladis Rodrguez, Lidia Moreno, andVictoria Pinedo). Thanks to the families of the children whoparticipated. We also thank the other members of the project advisorycommittee (Margaret Bentley, Patrice Engle, Ruth Frischer,Jean-Pierre Habicht, Sandra Huffman, Jane Lucas, Homero Martinez,Gretel Pelto, and Ellen Piwoz) for their helpful advice in thedevelopment of the study.

References1 Fishman S, Caulfield LE, de Onis M. Childhood and maternal

underweight. In: Ezzati M, Lopez AD, Rodgers A, Murray CJL, eds.Comparative quantification of health risks: global and regionalburden of disease attributable to selected major risk factors.Cambridge, MA, USA: WHO/Harvard University Press, 2004: 3961.

2 National Institute of Statistics and Information. Demographic andhealth family survey, 2000. Lima: Instituto Nacional de Estadisticae Informatica, 2003: 16386.

3 Rogers BL, Rajabiun S, Levinson J, et al. Reducing chronicmalnutrition in Peru: a proposed national strategy. Lima: USAID,2001.

4 Lutter CK, Rivera JA. Nutritional status of infants and youngchildren and characteristics of their diets.J Nutr2003;133: 2941S49S.

5 Freire W. Strategies of the Pan American HealthOrganization/World Health Organization for the control of irondeficiency in Latin America. Nutr Rev 1997; 55: 18388.

6 World Health Organization. Complementary feeding in developing

countries: a review of current scientific knowledge.WHO/NOT/98.1. Geneva: WHO, 1998.



10/10

Articles

7 Allen L, Gillespie S. What works? A review of the efficacy andeffectiveness of nutrition interventions. ACC/SCN nutrition policy

paper 19, 2001: 2728.8 Caulfield LE, Huffman SL, Piwoz EG. Interventions to improve

intake of complementary foods by infants 612 months of age indeveloping countries: impact on growth and on the prevalence ofmalnutrition and potential contribution to survival. Food Nutr Bull1999; 20: 183200.

9 Schroeder DG, Martorell R, Rivera JA, Ruel MT, Habicht JP. Agedifferences in the impact of nutritional supplementation ongrowth.J Nutr1995; 125 (suppl 4):1051S59S.

10 Guldan GS, Fan HC, Ma X, Ni ZZ, Xiang X, Tang MZ. Culturallyappropriate nutrition education improves infant feeding andgrowth in rural Sichuan, China.J Nutr2000; 130: 120411.

11 Bhandari N, Bahl R, Mazumdar S, Martines J, Black RE, Bhan MK.Infant feeding study group. Effect of community-based promotionof exclusive breastfeeding on diarrhoeal illness and growth: acluster randomized controlled trial. Lancet 2003; 361: 141823.

12 Webb R, Fernandez Baca G. Peru in numbers: 1999. Lima: CuantoS A, 1999: 129.

13 Lopez de Romaa G, Brown KH, Black RE. Health and growth ofinfants and young children in Huascar, Peru. Ecology Food Nutr1987; 19: 21329.

14 Dewey KG, Peerson JM, Heinig MJ, et al. Growth patterns ofbreast-fed infants in affluent (United Sates) and poor (Peru)communities: implications for timing of complementary feeding.Am J Clin Nutr1992; 56: 1011218.

15 World Health Organization. Medicin del cambio del estadonutricional. Geneva: WHO, 1983.

16 Creed-Kanashiro HM, Brown KH, Lpez de Romaa G, Lpez T,Black RE. Consumption of food and nutrients by infants inHuascar, an underprivileged community on the periphery ofLimaPer. Am J Clin Nutr1990; 52: 9951004.

17 Instituto de Investigacin Nutricional. Food composition tablesIIN, 2001. Lima: Instituto de Investigacin Nutricional, 2001.

18 Dewey KG, Brown KH. Update on technical issues concerningcomplementary feeding of young children in developing countries

and implications for intervention programs. Food Nutr Bull 2003;24: 528.

19 Joint FAO/WHO expert consultation. Vitamin and mineralrequirements in human nutrition. Geneva: WHO, 2002.

20 National Center for Health Statistics. Growth curves for childrenbirth18 years. Rockville: National Center for Health Statistics,1977.

21 Lubrook J, Dudley H. Why permutation tests are superior to t andF tests in biomedical research. Am Stat 1998; 52: 12732.

22 McNelly B, Dunford C. Impact of credit with education on mothersand their young childrens nutrition: lower Pra rural bank creditwith education program in Ghana (research paper no 4). Davis, CA:Freedom from Hunger, 1998.

23 Schroeder DG, Pachon H, Dearden KA, et al. An integrated childnutrition intervention improved growth of younger, moremalnourished children in Northern Viet Nam. Food Nutr Bull 2002;23: 5361.

24 Kramer MS, Chalmers B, Hodnett ED, et al. Promotion ofBreastfeeding Intervention Trial (PROBIT): a randomized trial inthe Republic of Belarus.JAMA 2001; 285: 41320).

25 Bhandari N, Mazumder S, Bahl R, Martines J, Black RE, Bhan MK.An educational intervention to promote appropriatecomplementary feeding practices and physical growth in ruralHaryana, India.J Nutr2004; 134: 234248.

26 World Health Organization, Divisin of Child Health andDevelopment. Integrated management of childhood illness.Document ref WHO/CHD/97.3. Geneva: WHO, 1997.

27 Santos I, Victora CG, Martines J, et al. Nutrition counselingincreases weight gain among Brazilian children.J Nutr2001;131: 286673.

1872 www thelancet com Vol 365 May 28 2005

Documents

IIN PennyMEetal.lancet.vol365