Diet and growth in infancy relationship

Availability of data

As a longitudinal study with many different outcomes measured in different ways at various ages, sometimes, in selected subsamples only, ALSPAC had inevitable varia- tions in the numbers of subjects and the level of missing data in each analysis. The number of subjects included in the main analysis for each paper, as well as whether the analysis used the CIF substudy only, is noted in the present review. In general, tables in the published papers showed the background biases inherent in studying the group with available data compared with the group without data for that particular analysis. These have not been reprinted in this review; they show biases similar to

those shown in Table 2 and these become more pro- nounced as the study progresses. ALSPAC has invested considerable effort into minimizing study attrition and has been reasonably successful. 67,68 An investigation into the effect of selective attrition in ALSPAC on behavioral outcomes was carried out using teacher-rated assessment of 8-year-old children in schools in the study area. 69 These data covered children currently engaged in the study (n = 3,975), those recruited but no longer taking part (n = 1,140), and those who had never been part of the study (n = 4,383). Simulations (n = 36) were carried out to explore the impact of selective recruitment and attrition. The analysis confirmed the selective bias in recruitment and attrition as shown in Table 2 and showed that behav- ioral problems were less prevalent in the retained group. However, the simulations showed that the prediction of teacher-reported behavioral problems by background factors was very similar in all three groups. The authors concluded that the presence of substantial selection bias did not greatly attenuate the relationship between expo- sure and this particular outcome, although there was a reduction in the range of the variables and an underesti- mation of prevalence of the outcome. Other outcomes may be affected more significantly.

Dietary assessment

Table 3 lists the ages at which dietary assessment was carried out and the numbers of subjects with data avail- able at each age.

Food frequency questionnaires. Maternal diet was assessed by an unquantified food frequency question- naire (FFQ) sent to the mothers at 32 weeks’ gestation. 70 The questionnaire contained questions about the weekly frequency of consumption of 43 foods and food groups. The mothers were asked to tick one of the following fre- quency options for each food: never or rarely, once in 2 weeks, 1–3 times per week, 4–7 times per week, or more than once a day. More detailed questions were asked about the consumption of a further 8 foods usually con- sumed daily (for example, bread, coffee, tea, and sugar). There were also questions about the types of certain foods used (for example, cooking and spreadable fats, milk, bread, and soft drinks) and about the ways in which foods were prepared and eaten, (for example, whether some or all of the fat was cut off meat, how often food was fried, and how many of the slices of bread eaten in a day were spread with fat). No questions were asked about portion

sizes; therefore, standard portion sizes 71 were used for the nutrient estimations. Questions about alcoholic drinks were asked, but they were not included in the analysis of nutrients. Nutrient intakes were calculated based on the

frequency with which each food was consumed. 70 Similar

4 Nutrition Reviews®

Nutrition

Reviews® Table 2 Socioeconomic background of mothers recruited to ALSPAC and to the intensively studied Children in Focus (CIF) 10% substudy of ALSPAC, along with

socioeconomic background of the mothers in both groups who supplied dietary information about their children in the second year of life.

Characteristic

Whole ALSPAC

Mothers who completed

Mothers of infants

Mothers of infants in CIF

cohort (n = 14,541)

infant FFQ at 15 mo

in CIF substudy

substudy who completed diet

(n = 11,025)

(n = 1,432)

record at 18 mo (n = 1,026)

Percent Maternal education No school qualification at age 16

21.1 School qualification obtained at age 16

35.9 Education beyond age 16

40.8 Data missing

3.9 66 4.6 22 2.2 Maternal age at birth <20 y

11.4 Data missing

Housing tenure Mortgaged or owned

80.7 Council and housing associations

9.9 Privately rented or other

9.3 83 8.1 Data missing

3.0 41 3.0 13 1.3 Ethnicity White

2.0 31 2.1 20 2.0 Data missing

4.3 64 4.5 22 2.1 Smoked in last trimester of pregnancy Yes

78.2 Data missing

9.8 Maternal prepregnancy BMI <18.5

4.9 86 6.0 58 5.7 Data missing

11.1 86 8.4 Abbreviations: BMI, body mass index; CIF, Children in Focus substudy; FFQ, food frequency questionnaire.

Table 3 Dietary data available in the ALSPAC recruited mothers' cohort (n = 14,541 a ), the ALSPAC children's cohort (n = 14,062 b ), and the CIF substudy (n = 1,432).

Age of study child

Sample

Dietary method used

Response rate No.

Percent

32 wks gestation

ALSPAC (mothers)

Infant feeding questionnaire

Substudy CIF

1-day food record

Infant FFQ

Substudy CIF

3-day food record

Infant FFQ

71.6 a 604 had no live birth, 69 had unknown outcome.

18 mo

Substudy CIF

3-day food record

b 208 had multiple births.

parent-completed FFQs covering the child’s diet at 7 household measures. Parents were asked to record the use years of age were also collected.

of medicines and vitamins in the drinks section of the diary. Mothers of breastfed infants were asked to record

Infant feeding questionnaires. Parents were asked to com- the duration of each feeding in minutes. plete questionnaires about their child’s health and devel-

Food records were converted into weights and codes opment

(http://www.bristol.ac.uk/alspac/researchers/ linkable to the energy and nutrient content correspond- resources-available/data-details/questionnaires).

ing to each of the drinks or foods consumed, using questionnaires were sent when children were ages 4

The

Microdiet software (University of Salford, UK) at 4 and 8 weeks, 6 months, and 15 months; all included a section

months and DIDO software (developed by the Dunn about the child’s diet. The questionnaire for 4-week-olds

Nutrition Unit, Cambridge, UK) 72 at 18 months. Portion collected data about breast and formula feeding, feeding

sizes for baby foods were usually described by spoonfuls schedule, feeding behaviors, and feeding difficulties. In

or by the proportion of a jar eaten. Family foods were the questionnaires about 6- and 15-month-olds, there

assigned weights based on reported food-portion sizes 73 were questions about breast and formula feeding, fre-

and scaled down appropriately. Weights were also quency of consumption of specific foods and drinks

obtained from manufacturer’s information and from (listing foods likely to be offered to infants in the UK), age

food packages. Composite foods and recipes that did not at the introduction of each food/drink, feeding behaviors,

have an equivalent in the food tables were broken down and feeding difficulties (refer to Table 3 for response

into their component parts and assigned weights rates). In the questionnaire about 15-month-olds,

appropriately.

mothers were asked, “Babies’ first solid meals are usually The data bank used for nutrient analysis of the food

a puree. When did your child first have a meal with records included the fifth edition of McCance and ‘lumps’ in?” This was to ascertain when chewy/lumpy

Widdowson’s The Composition of Food, 74 along with its foods were first introduced. None of the ALSPAC infants

supplements. 75–83 These data were used to generate received formula with added long-chain fatty acids

average daily nutrient intakes and the amounts of various because there was minimal availability of this type of

groups of food/drink consumed. Intakes from dietary formula in the early 1990s.

supplements were not included in the nutrient calcula- tions. A measure of added-sugars (nonmilk extrinsic

Diet records. As part of CIF, the infant’s diet was assessed sugars) intake was calculated by deducting the sugars

from milk, fruits, and vegetables from total sugars. Fiber Table 3 for response rates). The parents were invited, by

by a research clinic at 4, 29 8, 31 and 18 38 months (refer to

was measured as nonstarch polysaccharide (NSP); this post, to record in a structured diary all foods and drinks

type of fiber does not include resistant starch or lignin their child consumed over 3 individual days (1 day only at

substances, which are a part of fiber measured by other

4 months): preferably 1 weekend day and 2 weekdays. The

methods.

parents could choose which days to record, so days were For breast milk, the duration of each feeding was not necessarily consecutive. They were asked to bring the

used to estimate the likely volume of milk consumed. The completed food record to a research clinic where, at 8 and

basis of the calculation differed at each age and used

18 months, they were interviewed briefly by a member of validated assumptions. 84,85 At 4 months, 125 g was the nutrition team to check for completeness and to

allowed for a feeding lasting 10 minutes or longer, and a clarify any uncertainties in the diaries, such as cooking

percentage of 125 g was allowed for shorter feedings (i.e., methods. Foods and drinks consumed were recorded in

12.5 g of breast milk/min). At 8 months, a feeding lasting

Nutrition Reviews®

10 minutes or longer was assumed to deliver 100 g in amount, with a proportion of this assumed if the feeding was of shorter duration (i.e., 10 g of breast milk/min).

Reference nutrient values. The calculated nutrient intakes were compared with age-specific reference values recom- mended by the UK Department of Health’s Report on

Health and Social Subjects, no. 41. 86 Reference Nutrient

Intakes have several levels: the Estimated Average Requirement is the amount of a nutrient at which about half the population will need more and half the popula- tion less than this amount; the Reference Nutrient Intake is the amount of a nutrient sufficient for 97% of the popu- lation; and the Lower Reference Nutrient Intake (LRNI) is the amount that is adequate for only around 2.5% of the population.

Anthropometric measurements

Birthweights were obtained from the medical records, and supine length was measured with a Harpenden neonatometer (Holtain Ltd, Crosswell, Dyfed, UK) soon after birth by a member of the ALSPAC study team. The infants in CIF had their weight and height measured at the research clinics at 4, 8, 12, 18, 25, 31, 37, 43, 49, and 61 months. For the whole ALSPAC cohort, body weight and length at around ages 2, 9, and 19 months were available from routinely collected measurements performed by trained health professionals as part of the infant health surveillance program and were extracted from the local child health database. Later height and weight measure- ments for the whole cohort were obtained at the research clinics at ages 7, 9, 10, 11, 12, and 13 years.At these ages, fat mass was also assessed by bioelectrical impedance using a Bodystat 1500 impedance monitor (Bodystat Ltd, Isle of Man, UK). A Lunar Prodigy dual-energy X-ray absorptiometry (DXA) scanner (GE Medical Systems, Madison, WI, USA)was used to measure body composi- tion at ages 9, 11, and 13 years and provided estimates of total fat mass, lean body mass, and bone mass. Whole- body scans were carried out with children wearing light clothing, with metal objects such as bracelets and watches removed.

From these measurements of height or length (cm) and weight (kg), ponderal indices in infancy (weight/

length 3 ) and BMIs in childhood (weight/height 2 ) were

calculated. Body weight, height/length, and BMI at each time point were converted to standard deviation scores (SDS; sometimes known as “z scores”) by comparing

values with the British 1990 growth reference, 87 using

gestational age or the actual age at measurement for each individual. Weight gain was assessed by calculating the difference in z scores between weight at the beginning and at the end of the period, adjusting for regression

toward the mean using correlates from the British 1990 growth reference. 88 Using this increasingly preferred method, 54,89 weight gain becomes “conditional” on gender, age, and initial weight, thereby providing a more accurate measure of infant growth.

Hemoglobin and ferritin values

At the 8-, 12-, and 18-month visits to CIF research clinics,

a heel-prick capillary blood sample was collected. The hemoglobin concentration was assayed using the HemoCue B-Hb photometer (HemoCue Ltd, Derbyshire, UK). The blood was then centrifuged and the plasma removed and frozen. Plasma ferritin was assayed using a dissociation-enhanced lanthanide fluoroimmunoassay (DELFIA), a time-resolved fluoroimmunoassay system.

Blood pressure measurements

Blood pressure was measured at a research clinic when the children were 7.5 years of age. Systolic blood pressure and diastolic blood pressure were measured using a Dinamap 9301 Vital Signs Monitor (Morton Medical, London, UK). Using a cuff size appropriate for a child’s upper-arm circumference, two right-arm measurements were taken and the average recorded.

Girls' age at menarche

Data on whether the girls had started menstruating were collected at a clinic between February 2004 and Decem-

ber 2005. 62 Girls were asked during a measuring session in a private room if they had started menstruating, and if so, when. In total, 3,751 girls attended the clinic. Mean age of girls who attended the clinic was 12.89 (SD, 0.23) years, ranging from 11.30 to 14.34 years. Data on whether menarche had been reached were available from 3,298 girls, of whom 1,637 (50%) stated they had not. Of the 1,661 who had reached menarche, age at first period was available for 1,550; of these, 1,328 were white girls from singleton births.

IQ and child development data

Developmental data were obtained from questionnaires mailed to parents when their children were 6 months and

18 months of age. The questionnaires contained a series of questions assessing the mental and physical develop- ment of the child (taken from the Denver Developmental Screening Test 90,91 ); this provided a developmental score for each child. When children were 8 years of age, their IQ was measured at a research clinic. IQ was measured using

a validated age-adjusted shortened version of the Wechsler Intelligence Scale for Children. 92

Nutrition Reviews®

Maternal socioeconomic and anthropometric factors

assess the likely relevance of ALSPAC results to infant feeding practices in 2011. The DNSIYC used the same

Maternal age at delivery was calculated by subtracting the method as ALSPAC to estimate breast milk intake from the mother’s date of birth from the child’s date of birth. Infor-

duration of each breastfeeding. For 90 mother-child pairs, mation on highest maternal educational attainment was

the DNSIYC was able to use a stable isotope method to derived from a questionnaire sent out at 32 weeks’ gesta-

measure breast milk intake for comparison with the diary tion. Information on all the educational qualifications of

estimates. 94 The agreement between the two methods was the mother was obtained, and from this a 5-point scale of

considered moderate (r = 0.599), with the diary method highest educational attainment achieved was created,

found to be reasonably reliable for estimating breast milk with the following categories: no academic qualifications;

intake in the field.

vocational training (hairdressing, catering, etc.); O-level academic examination, usually taken at 16 years, or

Statistical methods

equivalent; A-level academic examination, usually taken at 18 years, or equivalent; and university degree. In some

Since this review features a host of papers, it is not prac- analyses, these were contracted to three categories: no

tical to list all the statistical methods used in individual academic qualification, or vocational training only;

papers. Descriptive statistics were provided within most O-level qualification; and A-level or degree qualification.

papers. Confidence intervals (CIs) are shown at the 95% Information on maternal smoking status was assessed by

level in all cases. Tests for associations between variables questionnaires sent during pregnancy and infancy.

were performed using various statistical tests, paired t Housing tenure data were also collected during preg-

tests, analysis of variance, and χ 2 test. Regression analyses nancy (types of housing: owned or mortgaged housing;

were carried out after adjustment for major confounders, rented council housing; privately rented housing). Data

depending on the outcome of interest, and were on financial difficulties, including difficulty affording

described in the text for each study. The main confound- food, were derived from questionnaires completed during

ing variables used in the various analyses were the indi- pregnancy and during infancy of the child. Mothers were

cators of socioeconomic background listed in Table 2. asked to report their prepregnancy weight and height by

questionnaire during pregnancy; these values were used to calculate maternal prepregnancy BMI. The mother’s

RESULTS

partner was also asked to report his or her height and weight by questionnaire.

Diet and feeding practices in early infancy Comparison of nationally representative

Infant milk-feeding practices were assessed by question-

dietary studies

naire at 6 months (n = 10,474). Exclusive breastfeeding (defined as breast milk with no other food or drink except

Several nationwide surveys of diet in infants and very water) was carried out by 55% of the cohort in the first young children are available and have been quoted in

month, 44% in the second month, and 31% in the third comparison with ALSPAC data in this review. The IFS,

month, with a further 24% feeding breast milk in combi- carried out every 5 years, has comparable questionnaire

nation with other foods/drinks at 3 months of age. 24 By data from 1990 to 2010 on breastfeeding and complemen-

the sixth month, only 36% of the infants were receiving tary foods. 21–23,93 The Food and Nutrient Intakes of British

any breast milk, always along with other foods. Infant

formula was used from birth in 22% of infants, with this nationally representative infants aged 6–9 months (n =

Infants Aged 6–12 months 84 collected food records from

percentage gradually increasing to 63% in the sixth 258) and 9–12 months (n = 230) in 1985–1986, fairly close

month. 24 These findings were similar to results from the to the time of the infancy of ALSPAC children. This study

IFS carried out in the early 1990s, when the prevalence of helped to inform the methods used in dietary assessment

breastfeeding at 6 weeks was 44% and at 6 months was in ALSPAC and provided comparable nutrient and food

22%. 22 These data are still comparable with more recent group information. A new national survey was published

data from the IFS 2010, in which the prevalence of in 2013, the Diet and Nutrition Survey of Infants and

breastfeeding at 6 weeks was 55% and at 6 months was

Young Children (DNSIYC 2011) 94 ; it collected 4-day

weighed food records, as well as blood samples and The infancy diet in ALSPAC was assessed by ques- anthropometric measures, cross-sectionally from 329 4-

tionnaire at 6 and 15 months (n = 6,065 with complete to 6-month-olds, 630 7- to 9-month-olds, 449 10- to

data). These data documented the start of the comple- 11-month-olds, and 1,274 12- to 18-month-olds from

mentary feeding process and were compared, by applying January to August in 2011.It has been used in this review to

a scoring system, with infant feeding guidelines (duration

Nutrition Reviews® Nutrition Reviews®

keted for infant consumption at that time. The main asso- of foods and drinks to introduce) compiled from inter-

ciation with the feeding of fruit drinks at 4 months was the national sources. 25 The score comprised 14 components

absence of older siblings in the family (P < 0.0001), with (the main ones are covered in this paragraph) and was

maternal education much less important (P < 0.05).Herbal developed to assess the degree of adherence to comple-

drinks were more likely to be fed by younger mothers and mentary feeding guidelines. The difference between the

those who had never breastfed (P < 0.0001). 27 There was lowest (furthest from the guidelines) and the highest

some evidence that the volume of milk drunk,either breast (closest to the guidelines) quintiles, derived from

or formula, was slightly lower if other drinks were also guidelines-adherence scores, for ALSPAC infants was

given (around 70 g to 114 g less), which implies a slightly most marked in breastfeeding duration (an average of 1

lower nutrient intake in these groups because these other month vs. 8 months, respectively; P < 0.001), feeding on

drinks have very little nutritive value. 28 demand (22% vs. 56%; P < 0.001), age at introduction of

Further evidence of social inequality in infant cow’s milk (9 months vs. 12 months; P < 0.001), number

feeding practices came from the diet diary at 4 months of portions of vegetables (1.0 vs. 1.7 per day; P < 0.001),

(n = 933), 29 the age at which the introduction of comple- fruit (0.7 vs 1.1 per day; P < 0.001), and servings of com-

mentary foods was recommended at that time. 19,20 mercial infant food (14 vs 8 per week). Better adherence

Marked differences were found between formula-fed and to the guidelines was more likely with higher maternal

breastfed infants in the age at introduction to foods/ age (P < 0.001) and higher maternal education (P <

drinks and in the types of foods used. 29 Formula-fed 0.001). Mothers who were overweight before pregnancy

infants were more likely to be eating complementary were less likely to follow the guidelines than those with a

solids (93%), particularly commercial infant foods, but healthy weight (P < 0.001) before pregnancy. In longitu-

there was no difference in milk volume between formula dinal analysis (n = 4,326), better adherence to the guide-

drinkers fed solids and those not fed solids. Around lines – adjusted for socioeconomic background – was

80% of breastfed infants were receiving complementary independently associated with increasing scores on a

solids, and those that were tended to consume slightly less “health-conscious” dietary pattern (β = 0.18; CI: 0.14,

milk than those totally breastfed. However, the average

0.21) and with decreasing scores on a “processed” dietary weight of breastfed infants who received solids was pattern (β = −0.16; CI: −20, −0.13) derived from the child

similar to that of exclusively breastfed infants at 4 months FFQ at age 7 years by principal components analysis. 26 and was comparable to that of formula drinkers. 29 The These results suggest that a good dietary foundation may

overall nutrient intake of solids-eating breastfed infants set the trend for later healthy eating patterns.

did not appear to be compromised by the intake of solid The drinks offered to infants, including breast milk,

foods, although it is possible that lower amounts of these were investigated in detail at 4 months, using the diet

nutrients were absorbed. 29

diaries in the CIF subsample (n = 993). 27,28 The strongest

A further analysis related recorded diet at 4 months determinant of breastfeeding was maternal education: a

to growth up to 5 years of age (n = 881 full-term single- quarter of the lowest education group (no qualifications),

tons). Formula-fed and mixed-fed (breast and formula) half of the moderately high group (those with academic

infants were assessed separately from breastfed infants. 30 qualifications at 18 years), and three-quarters of the

In the formula/mixed-fed but not the breastfed group, highest group (those with degrees) were still breastfeeding

there was evidence that those introduced to complemen- at 4 months. Similarly, in IFS 2010, breastfeeding at 4

tary solids before 3 months of age had higher energy months was highest among mothers who left full-time

intakes at 4 months than those introduced later (P = education after 18 years of age (56%) compared with those

0.002). This was also true of first-born compared with

later-born infants (P = 0.01). Again, in the formula/ of the mother was also strongly independently related to

who left at or before16 years of age (22%). 21 In CIF, the age

mixed-fed group only, greater energy intake at 4 months breastfeeding, with older mothers much more likely to

was related to greater weight gain between birth and 1, 2 breastfeed (54% in mothers ≥30 years vs. 21% in mothers

or, 3 years of age and higher mean body weight and BMI ≤25 years). 27 Again in IFS 2010, breastfeeding rates at 4

at 5 years of age (P < 0.001). 30 These associations were months were lowest among mothers under the age of 20

not altered by controlling for maternal age or maternal (15%) and highest among mothers aged 30 and over

education.

(53%). 21 Apart from breast and formula milk, fruit drinks

were fed to 24% and herbal drinks to 15% in CIF. 28 Fruit

Diet and feeding practices in mid-infancy

drinks were any fruit-flavored drinks,including fruit juice, dilutable drinks,and drinks marketed for infant consump-

Food/drink records from the CIF substudy (n = 1,131) tion. Herbal baby drinks were flavored drinks containing

were available again at 8 months, at which point all infants Nutrition Reviews® were available again at 8 months, at which point all infants Nutrition Reviews®

iron would be advantageous. This use of cow’s milk as a were very similar to those from the national survey by

breastfed at least once a day. 31 Food and nutrient intakes

main drink before 12 months of age is an ongoing

problem in the United Kingdom, as shown by the 5-yearly exactly the same proportion of breastfed infants at 7–9

Mills and Tyler 84 and with the DNSIYC 2011, 94 which had

IFS (8% of 9-month-olds in IFS 2000, 6% in IFS 2005, and months (23%) as ALSPAC at 8 months. Nutrient intakes

4% in IFS 2010). 13,21,93 ALSPAC has the advantage of being were adequate, with zinc and vitamin D being the nutri-

a longitudinal study with data on many aspects of ents most likely to cause concern. In Table 4, the types of

feeding, and in another analysis, mothers feeding formula foods eaten at 8 months by ALSPAC infants are compared

to infants at 4 months were found to be twice as likely with those eaten in six countries around the world by

as those feeding breast milk to have switched to cow’s infants taking part in a WHO study designed to produce

milk as the main drink by 8 months (17% vs. 8%, respec- optimal growth data for breastfed infants on which

tively). 27 The mothers who did this were more likely to

have vocational education with no academic qualifica- ALSPAC children consuming each of the food groups was

growth charts could be reliably based. 95 The proportion of

tions at 16 years, to live in council-rented housing, to within the range between 6 and 12 months shown for

have several children, and to find it difficult to afford most of the countries and was particularly similar to con-

food. 27 Therefore, this particular type of adverse feeding sumption levels in Norway and the United States. 95 practice is related to socioeconomic background in this

A further analysis of these diet diary data (n = 928

population.

full-term singletons) focused on differences in diet When dietary salt (sodium) intake in CIF children according to the type and volume of milk fed. It was

was investigated at 8 months 33 and compared with the found that 12.9% of mothers were feeding cow’s milk as

recommendation for this age group, which is not more the main drink, and many infants were fed more than

than 400 mg of sodium per day, 86 70% of the diets were

found to exceed this maximum. 33 Dietary sodium content recommendations suggest a limit of 500–600 mL per day

600 mL of milk (any type) per day. 32 The infant feeding

was not related to maternal education but was higher in

infants introduced early to complementary foods. Fur- breastfed infants, this is equivalent to no more than 6

for formula-fed infants at 6 months 20 ; if applied to

thermore, the use of cow’s milk as a main drink was one breastfeedings per day. Table 5 shows intakes of energy

of the contributors to excessive sodium intake, since and selected nutrients according to the type and volume

cow’s milk has a much higher sodium content than either of milk fed. 32 In both breastfed and cow’s-milk-fed

breast milk or formula. 76 Bread and breakfast cereals were infants, the nutrient profile was worse in those who con-

large contributors to sodium intakes. This suggests that sumed high volumes of milk. The breastfed infants

reformulating these products to reduce their sodium seemed to be able to regulate their energy intake, so that

content would be beneficial for infants. There was also those fed breast milk more than 6 times per day obtained

inappropriate use of salty flavoring (given to 23% of these less energy from solids, with a total energy intake similar

infants), including the use of gravy granules. 33 to that of breastfed infants who received fewer feedings. This was at the expense of the other nutrients, particularly

Age at which different types and textures of food

iron, which resulted in 41% having intakes of iron below

were introduced

the LRNI, an amount adequate for only 2.5% of the popu- lation. However, although infants fed more than 600 mL

The association between later diet and age at which dif- per day of either formula or cow’s milk had lower intakes

ferent types and textures of foods were introduced to of solid foods than those fed lower volumes of milk, they

infants was investigated using parent-completed ques- did not fully compensate and had higher overall energy

tionnaires about feeding practices in infancy and up to 7 intakes. 32 Furthermore, the groups who were fed cow’s

years of age. In the infancy questionnaires, there were milk had a poor nutrient profile, particularly if fed more

separate questions about the feeding of ready-prepared than 600 mL per day: they had very low iron and vitamin

baby foods and home-prepared family foods, which were

C intakes and high calcium intakes, with a large propor- used to gain insight into how later diet was affected by tion having intakes below the LRNI for iron. The analysis

the way fruits and vegetables were introduced. 34 FFQ data showed that infants fed cow’s milk were much more likely

to assess diet were available at 7 years for 6,157 of the to be anemic at 12 months (27% with hemoglobin <110 g/

children assessed in infancy, with a bias toward higher dL, P = 0.006) than infants fed formula (15%), and that

socioeconomic status of those providing information higher calcium intake was associated with poorer iron

compared with those not providing it (P < 0.001). The status. 32 Underlining the problem with low iron intakes in

frequency of consumption of home-prepared fruits and breastfed infants was evidence showing they were vulner-

vegetables at 6 months was positively related to the fre- able to anemia (27%), suggesting that increased con-

quency of consumption of all eight types of vegetables

Nutrition Reviews®

Nutrition

Reviews®

Table 4 The percentage of infants in six countries eating each food group at 6 and 12 months of age from the WHO Multicentre Growth Reference Study Group

(1997–2003) compared with the percentage of ALSPAC infants eating each food group (assessed by food records collected at 8 months of age in 1993). a

Food group

Age

Percentage of children eating each food group Brazil (n = 69) Ghana (n = 228) India (n = 173) Norway (n = 159) Oman (n = 153) USA (n = 121) ALSPAC, at 8

mo (n = 1,178) Grains

Legumes & nuts

Milk & dairy products

6 mo 20.3 59.2 64.2 13.2 49.7 9.1 23.4

12 mo 75.4 76.3 89.6 82.4 83.0 86.0

Flesh foods

Vitamin-A-rich fruit & vegetables b 6 mo –

Other fruit & vegetables b 6 mo 73.9 11.4 52.6 23.9 39.9 34.6 85.7

Sweet beverages

Fats & oils

a Adapted from the WHO Multicentre Growth Reference Study Group (2006). 95 b In some countries, fruit and vegetables have been divided into vitamin-A-rich and other.

Table 5 Energy and selected nutrient intake at 8 months, according to the type and volume of milk fed, the

amount of energy from complementary solid foods, and the proportion of infants with very low iron intakes. a

Nutrient intake per day

Feeding group and no. of children

P value

≥6 Breastfeedings/day (n = 41) Total energy (kJ)

<6 breastfeedings/day (n = 101)

0.5 Energy from solids (kJ)

<0.001 Total protein (g)

24 20 <0.001 Total iron (mg)

6.2 5 0.04 Total calcium (mg)

0.03 Total vitamin C (mg)

79 74 0.6 Total NSP b 4.6 3.7 0.02

Iron <4.2 mg (LRNI c ) (%)

≥600 mL formula/day (n = 230) Total energy (kJ)

<600 mL formula/day (n = 602)

<0.001 Energy from solids (kJ)

<0.001 Total protein (g)

27 28 0.04 Total iron (mg)

9 10.7 <0.001 Total calcium (mg)

<0.001 Total vitamin C (mg)

0.02 Total NSP b 4.3 3.7 <0.001 Iron <4.2 mg (LRNI c ) (%)

≥600 mL cow’s milk/day (n = 69) Total energy (kJ)

<600 mL cow’s milk/day (n = 79)

<0.001 Energy from solids (kJ)

<0.001 Total protein (g)

33.8 40 <0.001 Total iron (mg)

6.3 6.2 0.8 Total calcium (mg)

<0.001 Total vitamin C (mg)

57 56 0.8 Total NSP b 4.7 3.8 0.01

29.1 26.1 0.7 Abbreviations: LRNI, Lower Reference Nutrient Intake; NSP, nonstarch polysaccharide.

Iron <4.2 mg (LRNI c ) (%)

a Reproduced with permission from Hopkins et al. (2007). 32 b A measure of the fiber content of the diet. c The amount of nutrient that will be adequate for only 2.5% of individuals.

and both types of fruit covered in the 7-year FFQ, whereas the age at which lumps were introduced showed any rela- associations with ready-prepared versions were some-

tionship to the types of foods fed at 15 months or to past times negative, particularly for green vegetables (all

or present feeding difficulties, as perceived by the reported associations P < 0.0001). There was also evi-

parents. 35 The majority of the children (72%) were intro- dence of a moderating effect of frequency of feeding

duced to lumps at between 6 and 9 months of age, home-prepared vegetables at 6 months on the relation-

although 11% were introduced before 6 months (early) ship between age of introduction and vegetable eating at

and 18% were introduced from 10 months onward (late).

7 years. 34 Thus, if children were introduced to home- Older mothers with better education were more likely to prepared vegetables after 5.65 months of age, a high fre-

have introduced lumps at the recommended age. Com- quency (intake on most days) of home-prepared

pared with the group who were introduced to lumps at vegetable feeding was associated with a high intake at 7

the recommended age (i.e., the “recommended group”), years, but a low frequency was not. Frequency of feeding

the early group were fed a greater variety of family foods was less important when home-prepared vegetables were

at both 6 months and 15 months, while fewer of the late introduced earlier. All the analyses were adjusted for

group were given a full range of solid foods at 6 months, socioeconomic background, breastfeeding duration, and

and they were less likely to be having family foods at 15 age of introduction to any solid foods. 34 months. 35 There was evidence that those introduced to Information about when “lumps,” i.e., chewy textured

lumps late were more difficult to feed at 15 months than foods, had first been given to the child was obtained, and

they had been at 6 months; this was not the case in the infants were grouped according to whether the recom-

other two groups. Those introduced late were almost mendation (introduction between 6 and 9 months) was

twice as likely as the recommended group to have definite followed. The initial analysis (n = 9,360) assessed whether

likes and dislikes at 15 months (odds ratio [OR] adjusted

Nutrition Reviews® Nutrition Reviews®

1.68, 2.16). 35 without any other additions is sufficient to increase From the FFQ data collected at 7 years, unadjusted

acceptance and consumption. 96

associations suggested that the group introduced to lumps late continued to be less likely than the recom-

Diet and feeding practices in the second year of life

mended group to eat many foods, particularly fruit and vegetables. 36 Associations adjusted for socioeconomic

When the infants were 18 months of age, in 1994, diet background, breastfeeding duration, feeding difficulties at

diaries were available for 1,026 infants in the CIF

6 months, and age at introduction to fruit and vegetables substudy. Again, intakes were adequate for most nutrients showed that the late group were more likely to eat less

except vitamin D, zinc, and iron. 38 Results were very than one portion of fruit (OR: 1.33; CI: 1.14, 1.55) and

similar to those of a British national survey of 1.5- to vegetables per day (OR: 1.21; CI: 1.05, 1.41), while the

4.5-year-old children carried out in 1992–1993 using early group was less likely to eat less than one portion

weighed diet records. 97 In CIF, there was little evidence

that food intakes had been under-reported at 18 months, dren introduced to lumps late were slightly more likely to

(OR for low vegetable intake: 0.70; CI: 0.57, 0.85). 36 Chil-

as only 4.9% of infants were identified as potential under-

be choosy with food at 7 years (OR adjusted: 1.16; CI: reporters of energy intake when their intakes were com-

1.05, 1.28) than those introduced at the recommended pared with energy requirements derived from a study time, and those introduced early were slightly less to be

using double-labeled water measurements of energy choosy (OR adjusted: 0.82; CI: 0.73, 0.93). 36 expenditure in young children. 98

These findings led to ALSPAC becoming part of a The drinks consumed by CIF 18-month-olds were European Union (EU) collaboration to examine critical

investigated further. Cow’s milk was the predominant periods during infancy, a partnership that aimed to for-

drink, given to 94%, while only 3.5% were given any mulate future dietary guidelines for children. This con-

formula and 2.4% any breast milk. 39 Fruit-flavored drinks sortium, named HabEat, brought together cohorts of

were very commonly used, with half of the children fed children from four countries: the United Kingdom

drinks labeled as either low-calorie or low-sugar and con- (ALSPAC, the longest-running of the four cohorts),

taining artificial sweeteners. There was a strong educa- France, Portugal, and Greece. Data for all cohorts include

tional bias in the use of these drinks: 61% of mothers with detailed information about infant feeding practices,

vocational but not academic qualifications versus 25% of infancy and childhood diet, and cultural variation of

mothers with degrees fed these drinks. Tea or coffee was eating patterns. These data were used in parallel analysis

given to 17% of the children, again with a very strong bias to investigate factors affecting fruit and vegetable eating

in maternal education (ranging from 33% of the least- in children. 37 Breastfeeding duration was longest in Por-

educated to 7% of the most-educated mothers). Only 20% tugal, and complementary foods were introduced earliest

of mothers fed plain water as a drink, and in this case the in the United Kingdom. Longer duration of breastfeeding

educational bias was toward the higher-educated mothers, was consistently associated with higher intakes of fruit

who did this more often, a bias that also held for the use of and, particularly, vegetables at 2 and 4 years of age across

orange juice (fed by 19% overall) and apple juice (9%). 39 the countries, after adjustment for equivalent social vari-

The type of vessel used for drinks was recorded and, ables. In the United Kingdom, these associations were

despite the recommendation that all fluids should be fed present throughout childhood. In the United Kingdom

by a feeding cup by 12 months of age, 20 a large number of and France, where maternal fruit and vegetable eating was

18-month-olds were still using feeding bottles (10% exclu- assessed by FFQ, the association with breastfeeding

sively, and 53% some of the time). Of the children of

mothers in the lowest education group, one-quarter used a suggest an independent effect of breastfeeding on fruit

remained robust to adjustment for this. 37 These results

bottle only, half used both a bottle and a cup, and one- and vegetable eating. The other cohorts, however, did not

quarter used a cup only. Of the children of degree- replicate the association found in the United Kingdom

educated mothers, 5% used a bottle only, 43% used both a between late introduction of vegetables and lower intake

bottle and a cup, and more than half used a cup only. The of vegetables in childhood, although this association was

amount of milk fed was related to the type of vessel used. present at all ages assessed by ALSPAC. This discrepancy

When a bottle was used exclusively, the average intake was may be due to the fact that the other cohorts did not

488 g/d, compared with 381 g/d when a cup was used differentiate between home-prepared and ready-

exclusively; there was no difference in the total amounts of prepared vegetables, shown to be a very important dis-

other drinks fed. 39 These data suggest that the use of a

feeding bottle at 18 months may promote excess intake of conducted experimental work to assess food exposure

tinction in ALSPAC. 34 The HabEat consortium also

cow’s milk at a time when milk intake should be gradually and acceptance in infants and young children. This

replaced by the consumption of a variety of foods. Nutrition Reviews®

The finding that diet at 18 months provided marginal very similar to those shown in the dietary patterns analy- amounts of iron prompted investigation of the relation-

sis performed using the Infant FFQ in the whole cohort, ship between diet and hemoglobin and ferritin as a

reported below. 42

measure of iron status in these children (n = 796 single-

tons with blood samples). 40 There were univariate posi-

Dietary patterns in infancy

tive associations between energy-adjusted intakes of iron and vitamin C and ferritin and negative associations

Further investigation of the questionnaires covering the between intakes of calcium and NSP (fiber) and ferritin,

feeding of the infants at 6 and 15 months was conducted although hemoglobin was positively associated only with

by examining patterns of associations between foods vitamin C intakes. In adjusted models, calcium intake

eaten. Four underlying patterns were found at each age. 42 retained a strong negative association (P < 0.001) and

There were particular foods with a high correlation with iron intake retained a weak positive association with

each pattern, and these were used to describe that pattern. ferritin (P = 0.049), while vitamin C intake retained a

At 6 months, the following patterns were identified: moderately strong positive association with hemoglobin

“home-prepared meat, vegetables, and desserts” (HM (P = 0.004). There was also evidence that hemoglobin

Traditional); “biscuits, sweets, crisps, fizzy drinks, and tea” levels were slightly higher in consumers of fruits or veg-

(Discretionary); “ready-prepared baby foods” (RM Baby etables than in nonconsumers (P = 0.013 and 0.044,

Foods); and “breastfeeding, raw fruit, and vegetables” respectively). In consumers, the amount of cow’s milk

(Breastfeeding). Three patterns at 15 months had very drunk was negatively associated with ferritin (P < 0.001)

similar correlations with the same foods as at 6 months:

the HM Traditional, the Discretionary, and the RM Baby very similar to the findings at 8 months, thus increasing

but was not related to hemoglobin. 40 These findings are

Foods patterns. A fourth pattern was obtained in the confidence in their plausibility. 32 15-month-olds that had high correlations with “herbs,

Whether nutrient and food group differences cheese, fish, nuts, legumes, raw fruit, and vegetables” between recorded diets at 18 months were independently

(Contemporary). At this age, only 4% of infants were still associated with maternal characteristics such as educa-

breastfeeding, and breast milk did not contribute sub- tion, age at the birth of the child, and smoking status was

stantially to any of the patterns.

Each child received a score on each pattern, and the strongly associated with maternal education: children of

investigated more closely. 41 Nutrient intakes were most

level of their score was associated with maternal charac- the least-educated mothers, when compared with chil-

teristics. 42 At both ages, high scores on the HM Tradi- dren of the most-educated mothers, had lower intakes of

tional pattern were more likely when the mother had NSP and total sugars and higher intakes of nonmilk

academic qualifications at 18 years but were not indepen- extrinsic (added) sugars, while intakes of most key

dently related to maternal age. Maternal obesity before micronutrients, particularly iron, were lower. This was a

pregnancy was associated with lower scores on this reflection of the many differences in the food groups

pattern at 6 months but not at 15 months. The RM Baby consumed: children of the least-educated mothers con-

Foods pattern was not strongly associated with maternal sumed “meat products,” “chocolate,” “crisps and snacks,”

education at either age; this pattern was less likely at 15 “white bread,” “fried potatoes,” “tea,” and “diet soft drinks”

months if older siblings were present in the family. The more often and “cheese,” “yogurt and fromage frais,”

Discretionary pattern was strongly negatively related to “wholemeal bread,” “fruit” and “fruit juice” less often than

maternal age and education at both ages: high scores were those of the most-educated mothers. Despite having a

much more likely if the mother was young and had low worse nutrient intake profile, these children were less

educational attainment. Maternal obesity before preg-

nancy and being in a family with older children were also dren of the youngest mothers versus children of the

likely to take vitamin supplements. 41 In the diet of chil-

associated with higher scores on this pattern. oldest, starch intake was higher and total sugars intake

In confirmation of other analyses in ALSPAC, the lower, but there were no important associations with

Breastfeeding pattern at 6 months showed a strong gra- other nutrients. Some independent associations with food

dient with maternal education: scores increased markedly and drinks were found: children with younger mothers

as educational attainment improved. 42 This pattern was were more likely to consume “crisps and snacks” and “tea”

more likely if older siblings were present and mothers and less likely to consume “buns, cakes, and puddings,”

were in the oldest age group but was less likely if mothers

were obese before pregnancy. At 15 months, the Contem- dren of smokers had lower intakes of NSP than those of

“biscuits,” “semi-skimmed milk,” and “fruit juice”. 41 Chil-

porary pattern showed some similarities to the nonsmokers, and the main differences in foods consumed

Breastfeeding pattern, with a strong gradient for maternal were less frequent intakes of “buns, cakes and puddings,”

education and age, but there was no relationship to the “wholemeal bread,” and “fruit.” These associations were

number of older siblings or maternal obesity. 42

Nutrition Reviews®

Dietary patterns from the questionnaires were inves- months (0.97 [CI: 0.49, 1.45], P ≤ 0.001) and with the tigated in relation to the nutrients derived from the diet

Contemporary pattern at 15 months (0.67 [CI: 0.07, 1.26], diaries collected 2–3 months later in the CIF substudy. 43 P = 0.029). The HM Traditional pattern was inconsistent

The associations pointed to several differences in nutrient in its association with IQ, being positively associated at profiles, suggesting that these patterns derived from foods

6 months but showing no independent association at 15 consumed also describe variations in the nutrient com-

months. 44

position of the diet. In particular, at both ages, high scores

A further attempt to assess the basis for these asso- on the Discretionary pattern were associated with higher

ciations was made by combining the scores on similar intakes of sodium but lower intakes of NSP and most

dietary patterns across each age, using multilevel mixed micronutrients and, thus, a less favorable nutrient profile.

models 45 and including IQ measured at both 15 years and The RM Baby Foods pattern was negatively associated

8 years of age. Four trajectories of dietary patterns from 6 with sodium intake, possibly due to strict legislation in

months to 2 years were developed for each child, and each United Kingdom about the amount of sodium allowed in

trajectory was characterized by different foods and

named accordingly. High scores on the “Healthy” trajec- tive association with iron and vitamin C intakes at both

ready-prepared baby foods. 43 This pattern also had a posi-

tory were correlated with breastfeeding at 6 months, high ages, but the profile of some nutrients, such as selenium

intakes of fresh fruit and vegetables, cheese, and herbs at and NSP was less favorable. High scores on the HM Tra-

15 and 24 months, and high intakes of fish and eggs at 24 ditional pattern at both ages were associated with higher

months. This “Healthy” trajectory was associated with a protein intakes, but higher energy intake was associated

β-coefficient on full-scale IQ points: 1.07 (CI: 0.17, 1.97), with this pattern at 6 months only, and other relationships