NUTRITION RESEARCH PAPER TOPIC: Obesity among adolescents

Supplement Article

Diet, growth, and obesity development throughout childhood in the Avon Longitudinal Study of Parents and Children

Pauline M. Emmett and Louise R. Jones

Publications from the Avon Longitudinal Study of Parents and Children covering diet, growth, and obesity development during childhood are reviewed. Diet was assessed by food frequency questionnaires and food records. Growth data were collected by routine measurements, and in standardized clinics, body fatness was assessed by bioelectrical impedance and DXA (dual-energy X-ray absorptiometry) scans. Diets changed dramatically during the preschool period with an increase in the intake of free (added) sugars (12.3% rising to 16.4% of energy) that remained similar until adolescence. This was due to increased intake of energy-dense, nutrient-poor foods. Two periods of rapid growth were identified; infancy and mid-childhood (ages 7–11 y) and both were associated with obesity development. Diets with high energy density were associated with increasing fat mass from mid-childhood until adolescence. Genetic and dietary factors showed independent associations with increasing adiposity. At all ages studied, there were dietary inequalities related to maternal educational attainment that may influence inequalities found in obesity development. The Avon Longitudinal Study of Parents and Children has provided valuable insights into how disparities in diet and growth may affect the development of ill health in adulthood.

INTRODUCTION

A balanced diet in childhood is very important to en-

sure optimum growth and development at a time of rapid growth. Poor diet has been associated with many

adult health conditions, such as coronary heart disease, diabetes, and some cancers.1–3 Establishing and main-

taining healthy eating habits is important because habits formed in early life are likely to continue into adult-

hood.4 The Bogalusa Heart Study found that coronary atherosclerosis and essential hypertension can begin in

childhood.5,6 Other work has shown that childhood fruit consumption may have a long-term protective

effect on cancer risk in adulthood.7 Therefore, an understanding of how dietary habits develop during

childhood and adolescence would be informative, espe-

cially for establishing critical points for intervention to prevent the development of later health problems.

Obesity is a chronic health condition that can man-

ifest itself well before adulthood, and there has been a rapid rise in the prevalence of childhood obesity in re-

cent years in the Western world.8–10 There is evidence that rapid weight gain in early childhood is predictive

of the development of obesity during childhood.11 It may be that early childhood is important because it is

when particular dietary habits and sedentary behavior patterns become established.12 In later childhood and

adolescence, promoting a healthy lifestyle with a bal- ance between diet and physical activity is essential to ar-

rest obesity development.13 In particular, the World

Affiliation: P.M. Emmett is with the Centre for Child and Adolescent Health, School of Social and Community Medicine, University of Bristol, Bristol, United Kingdom. L.R. Jones is with the School of Social and Community Medicine, University of Bristol, Bristol, United Kingdom.

Correspondence: P. Emmett, Centre for Child and Adolescent Health, School of Social and Community Medicine, University of Bristol, Oakfield House, Oakfield Grove, Bristol, BS8 2BN, UK. E-mail: [email protected]. Phone: þ44-117-3310195. Key words: ALSPAC, childhood diet, diet, energy density, fat mass, fruit and vegetables, growth, inequality, obesity, sugar.

VC The Author(s) 2015. Published by Oxford University Press on behalf of the International Life Sciences Institute. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/ licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited.

doi: 10.1093/nutrit/nuv054 Nutrition ReviewsVR Vol. 73(S3):175–206 175

Health Organization has identified the energy density

of diets and their fiber content as important factors for determining obesity risk.1

The Avon Longitudinal Study of Parents and Children (ALSPAC) has collected comprehensive die-

tary information and growth measures throughout childhood and has followed obesity development into adolescence. In this respect, it has a unique combina-

tion of longitudinal data collected by hands-on standardized procedures, as well as from parental or

self-completed questionnaires and routine community and health service assessments. This review aims to

amalgamate the publications that used ALSPAC data and covered childhood diet, growth, and obesity devel-

opment. It includes all published articles that examine food and nutrient intakes throughout the childhood of

the cohort and those that relate diet and growth to obe- sity and other markers of health. The relationship of

diet, growth, and obesity development with socioeco- nomic background (SEB) is explored. The diet and

growth data were collected from the same children at set ages ranging from preschool to mid–secondary

school age. Other reviews in the supplement of which this article is a part cover the pregnancy diet and the di-

etary patterns in the ALSPAC. A review of the ALSPAC’s contribution to the understanding of infancy

diet and growth is published elsewhere.14

LITERATURE AND STUDY METHODS

Literature

This narrative review includes all articles using data from the ALSPAC that deal with dietary intake, growth,

and obesity development between the ages of 1.5 and 15 years. The included articles are listed in Table 1. Of the

57 articles identified15–72 4 used dietary questionnaire data, 28 used food record data, 8 assessed growth, 32 as-

sessed obesity by body mass index (BMI) or fat mass, and 8 focused on diet and growth in relation to other

health outcomes.

Subjects

The ALSPAC is a birth cohort study that recruited preg-

nant women resident in 3 health districts surrounding the city of Bristol in the county of Avon, in southwest

England, with an expected delivery date between April 1991 and December 1992 (n ¼14 541 pregnancies).73,74 It was set up to investigate the ways in which genes and the environment, including diet, interact to affect the

health, behavior, and development of children. Ethical approval for the study was obtained from the ALSPAC

Law and Ethics Committee and the local research ethics

committees. The cohort was population based and

broadly representative, at recruitment, of the population of women with children aged <1 year in Avon.74 The

indicators of the SEB of the family at recruitment are shown in Table 2. The children (n¼14 062 at birth; n¼13 988 alive at 1 y) have been followed using ques- tionnaires completed by parents and the children, edu- cational records, and hands-on assessment at dedicated

research clinics.73 A proportion of children born in the last 6 months of the recruitment phase (equivalent to

10% of the whole cohort) was selected to take part in a substudy known as Children in Focus (CIF).73 Parents

of these children were invited to bring their child to re- search clinics at intervals from age 4 months to age 5

years (n¼1432 ever attended). When the children were aged 7 years, the whole cohort was invited to attend the

research clinic. At this age, an attempt was made to bol- ster the initial sample with eligible subjects who failed

to join the study originally; 713 additional children (from 706 pregnancies) were recruited (total child co-

hort: n¼14 701 alive at 1 y). The clinic evaluations were repeated annually until age 15 years. The ALSPAC

website contains details of all of the data that are available through a fully searchable data dictionary.75

Table 2 also shows the SEB of the mothers who com- pleted a food record for their CIF child at age 3 years

and the SEB of mothers of children who attended the research clinic and kept food records at ages 7 years

and 13 years in comparison with the originally recruited mothers. The retained mothers have higher education

attainment, are older, and have more favorable health indicators than mothers whose children did not com-

plete the follow-ups.

Dietary assessment

Table 3 shows the timing and response rates for the col- lection of the dietary data by food frequency question-

naire (FFQ) and food record.

Food frequency questionnaire

Maternal diet was assessed using an unquantified self-

completion FFQ sent to the mothers at week 32 of ges- tation.76 It was later adapted to cover the child’s diet

and was completed by parents about the child at various ages (Table 3). The parental questionnaire when the

child was aged 13 years asked only about foods pro- vided by the parents to the child; the adolescents them-

selves were asked to complete a separate questionnaire about foods they ate outside of the home, such as school

meals, take-away foods, confectionery, and soft drinks.77 For the estimation of energy and nutrient

intakes, the data from these questionnaires were

176 Nutrition ReviewsVR Vol. 73(S3):175–206

Table 1 Characteristics of the ALSPAC articles included in the present review Reference Dietary method

(sample) Age at assessment

Type of analysis

Focus of paper

Cowin et al. (2000)15 FR (ss) 1.5 y XS Diet description and adequacy Emmett et al. (2002)16 FR (ss) 1.5 y and 3.5 y XS and L Diet description and adequacy

Change in foods and nutrients Glynn et al. (2005)17 FR (ss) 7 y XS Diet description and adequacy

Sex differences Cribb et al. (2011)18 FR (AA) 10 y XS Diet description and adequacy

SEB differences Misreporting of energy intake (EI)

Anderson et al. (2013)19 FR and FFQ (AA) All ages 3–13 y XS and L Trajectory of EI by multilevel modeling Prepregnancy BMI and childhood EI EI and child BMI at 15 y

Rogers et al. (2003)20 FR (ss) 1.5 y XS SEB food and nutrient differences Rogers et al. (2002)21 FR (ss) 1.5 y and 3.5 y XS and L Fat intake quartiles association with foods and nutrients Rogers et al. (2001)22 FR (ss) 1.5 y and 3.5 y XS and L Fat intake quartiles association

with growth and obesity, iron status and blood lipids Cribb et al. (2013)23 FR (ss) 1.5 y and 3.5 y XS and L Vitamin A and carotene

Core and noncore foods Cribb et al. (2014)24 FR (ss) 1.5 y and 3.5 y XS and L Vitamin D and calcium

Fortification model Sayers et al. (2012)25 (AA) Mid-childhood Vitamin D status and cortical bone Johnson et al. (2007)26 FR (ss) 5 y and 7 y XS and L Dietary energy density associations

with nutrients; misreporting of energy intake; tracking of intake and association with fat mass at 9 y

Johnson et al. (2009)27 FR (AA) 10 y XS Energy density and FTO genotype association with fat mass

Jones et al. (2010)28 FR (AA) 7 y; Maternal FFQ XS Fruit and vegetable intake Johnson et al. (2007)29 FR (ss) 5 and 7 y XS and L Drinks description and SEB;

association with fat mass at 9 y Jago et al. (2010)30 FR (AA) 10 y; physical

activity 11 y XS Dietary association with physical activity

Noel et al. (2010)31 FR (red s) 13 y; physical activity 13 y

XS Use of measured physical activity for estimation of misreporting

More et al. (2014)32 FR (ss) 1.5 y and 3.5 y XS Portion sizes and foods consumed Rogers et al. (2007)33 FR (ss) 7 y XS School meals dietary intake Wright et al. (2008)34 (ss) Birth to 5 y XS Growth standards compared Reilly et al. (2000)35 (AA) 7 y XS Use of BMI to identify obesity Reilly et al. (2010)36 (AA) 9 y XS BMI and waist circumference association with fat mass

from DXA Sherriff et al. (2009)37 (AA) 7–11 y XS and L Fat and lean mass association with fitness and grip

strength Reilly et al. (2010)38 (ss) 11 y XS Validation of fat mass measurements Hughes et al. (2011)39 (ss) Birth to 15 y L Timing of excess weight gain Din et al. (2013)40 (AA) Birth to 13 y L Timing of periods of weight gain Rogers et al. (2006)41 (AA) 9 y L Birth weight and ponderal index association with fat and

lean mass at 9 y Howe et al. (2010)42 (AA) Birth to 15 y L Dponderal index and DBMI association with fat mass at

15 y Ong et al. (2009)43 (AA) Girls only, birth

to 9 y L Infancy weight gain association with BMI and fat mass,

overweight McCarthy et al. (2005)44 (ss) 2.5–5 y L Change in BMI and waist circumference Reilly et al. (1999)45 (ss) 4 and 5 y XS Obesity prevalence Hughes et al. (2011)46 (AA and ss) 3–15 y L Incidence of obesity Reilly et al. (2011)47 (AA) 7–13 y L Progression from overweight to obesity Wright et al. (2010)48 (AA) 7–11 y L Tracking of fatness and obesity Reilly et al. (2005)49 (AA and ss) 7 y XS and L Risk factors for obesity Sovio et al. (2011)50 (AA) Birth to 10 y MA FTO genotype and BMI Howe et al. (2012)51 (AA) Birth to 10 y L SEB differences in height trajectories Howe et al. (2010)52 (AA) Birth to 10 y L SEB differences in adiposity trajectory Howe et al. (2010)53 (AA) 9 y and 10 y XS SEB differences in CVD risk Dorosty et al. (2000)54 FR (ss) 1.5 y XS and L Adiposity rebound and diet, association with parental

obesity Timpson et al. (2008)55 FR (AA) 10 y XS Diet association with FTO gene

(continued)

Nutrition ReviewsVR Vol. 73(S3):175–206 177

combined. Maternal diet was assessed again 4 years af- ter the birth of the study child using the modified FFQ.

Copies of all the questionnaires used are available on- line.78 For foods/drinks not usually consumed every

day, the person was asked to indicate how often the food was consumed currently, using the following op-

tions: 1) never or rarely; 2) once in 2 weeks; 3) 1–3 times a week; 4) 4–7 times a week; and 5) more than

once a day. For frequently eaten foods, there were more detailed questions, such as how many cups of tea or cof-

fee were consumed, how much milk and sugar were used in each cup, the number of slices of bread con-

sumed each day, on average, and how many slices were spread with butter/margarine. Questions also covered

the usual type of milk (full fat or other), the usual type of bread (white, brown, or wholemeal), and the usual type of spread (butter/margarine or other) consumed.

To calculate the amount of each food consumed per week, the frequency questions were converted as fol-

lows: 1) 0; 2) 0.5; 3) 2; 4) 5.5; and 5) 10 times per week. Consumption of tea, coffee, bread, milk, and fat spreads

was recorded on a daily rather than a weekly basis. No questions were asked about portion sizes; therefore,

standard portion sizes79,80 tailored to the age of the per- son being assessed were used for the nutrient estima-

tions. Nutrient intakes were calculated based on the frequency with which each food was consumed and the

nutrient content of a portion of that food.76 To avoid including outliers with abnormally high or low intakes,

at each age subjects with very high or very low intakes

of energy were removed from the nutrient analysis after inspection of the histogram for energy intakes at that

age (between 220 and 330 subjects excluded).

Food records

Food records were collected 6 times between the ages 1.5 years and 13 years, as listed in Table 3. Diet was as-

sessed by parental completion of a food record in the CIF subsample at ages 1.5, 3.5, and 5 years15,16 and in

the whole ALSPAC cohort plus the new recruits at age 7 years.17 The parents were invited by post to record in a

structured record all foods and drinks their child con- sumed over 3 individual days (preferably 1 weekend

day and 2 weekdays, which did not need to be consecu- tive). They were asked to bring the completed food re- cords to the clinics, where, when the children were aged

1.5 years and 3.5 years, but not aged 5 years and 7 years, they were interviewed briefly by a member of the nutri-

tion team to check for completeness and to clarify any uncertainties in the records, such as cooking methods.

Foods and drinks consumed were recorded in house- hold measures. At ages 10 years and 13 years, food re-

cords were targeted for completion by the child with parental help.18 The child and parent were interviewed

during the clinic by a nutrition fieldworker for up to 15 minutes to obtain further details about the foods con-

sumed.18 If no diary had been brought to the clinic, the nutrition fieldworker carried out a 24-hour recall for

the previous day. A short questionnaire accompanied

Table 1 Continued Reference Dietary method

(sample) Age at assessment

Type of analysis

Focus of paper

Fraser et al. (2011)56 FFQ (AA) 13 y XS Frequency of eating fast food association with healthy foods and BMI

Noel et al. (2011)57 FR (red s) 10 y and 13 y XS and L Milk intake association with body fat percentage Noel et al. (2013)58 FR (red s) 10 y and 13 y XS and L Flavored milk association with change in body fat

percentage Bigornia et al. (2014)59 FR (red s) 10 y and 13 y XS and L Dairy food association with body fat percentage Cowin et al. (2001)60 FR (ss) 1.5 y XS Diet association with iron status Cowin et al. (2000)61 (ss) 2.5 y and 3.5 y XS and L Birth weight and size, obesity association with blood

lipids Cowin et al. (2001)62 FR (ss) 1.5 y XS and L Diet association with blood lipids at 2.5 y Rogers et al. (2006)63 (ss) 5 y and 7–8 y XS and L Height association with IGF axis Rogers et al. (2006)64 FR (ss) 7 y XS Milk and dairy association with IGF axis Rogers et al. (2005)65 FR (ss) 7 y XS Food and nutrient association with IGF axis Ong et al. (2004)66 (ss) 7–8 y XS and L Insulin sensitivity and secretion Salgin et al. (2012)67 (ss) 7–8 y XS and L Free fatty acids and insulin secretion Shultis et al. (2005)68 FR (ss) 1.5 y, 3.5 y, and

7 y L Diet association with birth weight

Rogers et al. (2010)69 FFQ and FR (AA) 3 y, 7 y, and 10 y L Diet in relation to age at menarche in girls Timpson et al. (2007)70 (AA) 10 y XS Genetics of bitter taste detection Golding et al. (2009)71 (AA) 4 y and 10 y L Growth and feeding behavior in bitter “super” tasters Emond et al. (2010)72 FFQ (AA) 3 y and 4 y L Diet and growth in autism spectrum disorder Abbreviations: AA, all available subjects from the whole cohort; BMI, body mass index; CVD, cardiovascular disease; DXA, dual-energy X-ray absorptiometry; EI, energy intake; FFQ, food frequency questionnaire; FR, food record; IGF, insulin-like growth factor; L, Longitudinal; MA, meta-analysis; red s, reduced sample size excluding FR with 1 and 2 days only; SEB, socioeconomic background; ss, CIF sub-sample; XS, cross-sectional; y, age of child in years D, change in

178 Nutrition ReviewsVR Vol. 73(S3):175–206

the food record and provided further details to aid in- terpretation of the food record. For example, the vol-

ume of the usual cup used for drinks was recorded, as well as information about any dietary supplements

used. The food records including the 24-hour recalls

were transformed into weights and codes linkable to the nutrient content corresponding to each of the drinks or

foods consumed using DIDO software, developed by the Human Nutrition Research Unit in Cambridge,

United Kingdom.81 Portion sizes for foods were de- scribed using household measures. Food weights were

allocated based on described portion sizes; if the de- scription was inadequate, portion size was based on

weighed intake data from a national sample of similarly aged children.80 Food weights were also obtained from

those given on packets. Composite foods and recipes that did not have an equivalent in the food tables were

broken down into their component parts. The data bank used for the nutrient analysis of the

food records included the 5th edition of McCance and Widdowson’s food tables82 and the supplements to the

tables.83–91 Nutrient information for any foods not cov- ered by this database was obtained from the National

Diet and Nutrition Survey database92 or from manufac- turers’ information. These data were used to generate

average daily nutrient intakes and amount consumed of various groups of food. Intakes from dietary supple-

ments were not included in the nutrient calculations. In some analyses only children with 3 complete days of

food records were included as stated in the text. The en- ergy density of the diet was investigated by dividing

Table 2 Socioeconomic background of the mothers recruited to the ALSPAC, including those who supplied dietary infor- mation at age 3 years for their child in the 10% subsample of ALSPAC, and those who supplied dietary information about their child at ages 7 years and 13 years Characteristic No. (%) of

recruited mothers (n¼14 541)

No. (%) of mothers who completed food record for child at age 3 years (CIF only; n¼863)

P-value No. (%) of mothers who completed food record for child at age 7 years (n¼7285)

P-value No. (%) of mothers who completed food record for child at age 13 years (n¼6112)

P-value

Level of education <0.001 <0.001 <0.001 No school qualifications

at age 16 y 3709 (25.5) 174 (20.2) 1390 (19.1) 1151 (18.8)

School qualification obtained at age 16 y

4273 (29.4) 308 (35.7) 2384 (32.7) 1982 (32.4)

Education beyond age 16 y 4358 (30.0) 363 (42.1) 2907 (39.9) 2499 (40.9) Missing 2201 (15.1) 18 (2.1) 604 (8.3) 480 (7.9)

Maternal age at birth (years) <0.001 <0.001 <0.001 <20 655 (4.5) 13 (1.5) 112 (1.5) 99 (1.6) 20–24 2682 (18.4) 90 (10.4) 911 (12.5) 745 (12.2) 25–29 5369 (36.9) 356 (41.3) 2718 (37.3) 2309 (37.8) 30–34 3808 (26.2) 304 (35.2) 2299 (31.6) 1943 (31.8) �35 1382 (9.5) 100 (11.6) 862 (11.8) 710 (11.6) Missing 645 (4.5) 0 383 (5.3) 306 (5.0)

Housing tenure <0.001 <0.001 <0.001 Mortgaged or owned 9757 (67.1) 700 (81.1) 5641 (77.4) 4742 (77.6) Council and housing association

rented 2138 (14.7) 80 (9.3) 584 (8.0) 467 (7.6)

Privately rented or other 1440 (9.9) 70 (8.1) 492 (6.8) 434 (7.1) Missing 1206 (8.3) 13 (1.5) 568 (7.8) 469 (7.7)

Ethnicity <0.001 <0.001 <0.001 White 11927 (82.0) 830 (96.2) 6549 (89.9) 5517 (90.3) Nonwhite 321 (2.2) 15 (1.7) 112 (1.5) 102 (1.6) Missing 2293 (15.8) 18 (2.1) 624 (8.6) 493 (8.1)

Smoked in last trimester of pregnancy

<0.001 <0.001 <0.001

Yes 2413 (16.6) 110 (12.8) 897 (12.3) 683 (11.2) No 8859 (60.9) 664 (76.9) 5254 (72.1) 4468 (73.1) Missing 3269 (22.5) 89 (10.3) 1134 (15.6) 961 (15.7)

Maternal prepregnancy BMI <0.001 <0.001 <0.001 <18.5 567 (4.0) 36 (4.2) 259 (3.6) 226 (3.7) 18.5–24.99 8564 (58.2) 584 (67.7) 4761 (65.4) 4029 (65.9) 25–29.99 1739 (12.0) 129 (14.9) 924 (12.7) 774 (12.7) �30 631 (4.3) 32 (3.7) 317 (4.4) 257 (4.2) Missing 3033 (20.8) 70 (8.1) 1024 (14.1) 826 (13.5)

Abbreviations: BMI, body mass index; CIF, Children in Focus.

Nutrition ReviewsVR Vol. 73(S3):175–206 179

total food energy (kJ) by the total weight (g) of food

consumed. The energy contribution from drinks was assessed separately. A measure of added or free sugars

intake (equivalent to non-milk extrinsic sugars) was cal- culated by deducting the sugars from milk, fruits, and

vegetables from total sugars.93 Fiber was measured as nonstarch polysaccharide; this type of fiber does not in-

clude resistant starch or lignin substances, which are part of the fiber measured by some other methods.93

For some analyses, foods were grouped into nutri- ent-rich core food groups (bread, rice, pasta; fruits; veg-

etables; meat, fish, eggs; milk, cheese) or nutrient-poor noncore food groups (cakes, biscuits, confectionery, savory snacks; processed meat/poultry/fish; potatoes/

vegetables with fat; spreads, sauces) according to guide- lines used in Australia.94 This was done to differentiate

foods that are recommended to be eaten by children as part of a balanced diet from foods that are extra to the

basic diet and tend to unbalance it. These noncore foods are sometimes referred to as discretionary foods.

Misreporting of energy intake

For the food records, misreporting of energy intake (EI) was assessed using an individualized method that takes

into account the age, sex, and body weight of the child and allows for growth and a standard level of physical ac-

tivity.95 The ratio of reported EI to estimated energy re- quirement (EER) was calculated (EI:EER). Individual

EERs were estimated using equations from an Expert Consultation Report on Human Energy Requirements.96 A

95% confidence interval (CI) for the accuracy of EI:EER was calculated by taking into account the amount of var-

iation inherent in the methods used to estimate EI and EER.97 For example, the confidence range for EI:EER,

calculated for the data at age 7 years, was 0.79–1.21, so reports of EI between 79% and 121% of EER were con-

sidered to be within the normal range of measurement

error and were defined as plausible reports. Those below

this cut-off were under-reports, and those above were over-reports. Slightly different cut-offs based on the age-

specific data were applied at other ages.

Energy adjustment

Energy adjustment is necessary when assessing relation- ships between nutrients or food groups and age-related

outcomes because energy is highly correlated with most nutrients and related to body size98; it may, thus, ob-

scure underlying relationships. The residuals method described by Willet has been used to adjust nutrient in-

takes.98 For food groups, the weight of food eaten has been divided by the total energy content of the diet at

each age so that intakes are directly comparable be- tween the ages and sexes. Energy adjustment provides

an assessment of the quality of the diet as opposed to the quantity provided by the unadjusted data.

Context for the ALSPAC data: reference nutrient intakes and national cross-sectional studies of diet in children

Dietary reference values for food energy and nutrients

for the UK population are used to assess the adequacy of the diet in the ALSPAC children at each age.93 There

is no published United Kingdom reference value rec- ommended for fiber (nonstarch polysaccharide) in

childhood; for adults, the recommendation is 18 g non- starch polysaccharide per day. The Caroline Walker

Trust has researched this matter99 and suggests that children should consume a percentage of adult intake

related to their EI. They suggest 8 g nonstarch polysac- charide/4.2 MJ (1000 Kcal) should provide an adequate

intake of fiber. In the United Kingdom, food-based rec- ommendations for healthy eating are based on the

Eatwell plate.100

Table 3 Dietary data available for the ALSPAC mothers during pregnancy (recruited, n¼14 541a), the ALSPAC children (at birth, n¼14 062b), and a 10% subsample of children who were studied more intensively (ever attended, n¼1432) with response rates Age of child Sample Type of dietary data Response (n) Rate (%) 32 weeks gestation Mothers FFQ 12 423 85.4 1.5 y 10% subsample 3-day food record 1026 71.6 3 y Children FFQ 10 137 69.7 3.5 y 10% subsample 3-day food record 863 60.3 4 y Children FFQ 9715 66.8 5 y 10% subsample 3-day food record 772 53.9 7 y Children FFQ 8505 58.5 7 y Children 3-day food record 7285 50.1 9 y Children FFQ 8223 56.6 13 y Children 3-day food record 6112 42.0 13 y Children FFQ 7079 48.7 a 604 no live birth, 69 unknown outcome. b 208 multiple births. Abbreviation: FFQ, food frequency questionnaire.

180 Nutrition ReviewsVR Vol. 73(S3):175–206

The rolling program of the National Diet and

Nutrition Survey (NDNS) carries out regular cross- sectional recording of diets as eaten by children aged

1.5–18 years in the United Kingdom using 4-day101 or 7-day92 weighed food records or food records described

using household measures102 as the assessment tools. The children aged 1.5–4.5 years who were surveyed by the NDNS in 1992/1993101 have been compared with

the ALSPAC children at ages 1.5 years and 3 years.15,16

The NDNS 199792 children have been compared with

the ALSPAC children at ages 7 years and 10 years.17,18

An update and extension of the NDNS in 2008–2012102

carried out over 4 years in children aged 1.5–18 years has been used to assess whether diet in children at dif-

ferent ages has changed since the time of the ALSPAC data collection.

Physical activity

The children were asked to wear an MTI actigraph

model 7164 (Manufacturing Technology Inc, Fort Walton Beach, FL, USA) for 7 consecutive days103 at

ages 11, 13, and 15 years to measure their physical activ- ity. Although these monitors cannot capture data for

some activities (swimming, contact sports, cycling), they have been shown to provide a reasonable estimate

of physical activity.104 The activity data collected were included in the analyses if there was >3 days of data

with at least 600 minutes of data per day.105 Mean counts per minute, an indication of the volume of phys-

ical activity, was calculated. Moderate to vigorous physi- cal activity was represented by the mean minutes per

day in which there were >3600 accelerometer counts per minute.

Anthropometric measurements

Birth weights were obtained from the medical records,

and supine length was measured soon after birth by a member of the ALSPAC study team. Body weight and length at approximately ages 2, 9, and 19 months and at

school entry were available from routinely collected measurements performed by health visitors as part of

the child health surveillance program and were ex- tracted from the local child health database.

Anthropometric measurements were taken at each research clinic visit using standardized procedures de-

scribed in each published article. In clinic visits from age 7 years, fat mass was assessed by bioelectrical im-

pedance using a Tanita leg-to-leg body fat analyzer (Model TBF 305; Tanita, Tokyo, Japan) and at age 7

years only by an arm-to-leg impedance monitor (Bodystat 1500; Bodystat Ltd., Isle of Man, UK). A

whole-body DXA (dual energy X-ray absorptiometry)

scanner (Lunar Prodigy DXA scanner; GE Medical

Systems, Madison, WI, USA) was used to measure body composition from age 9 years onward and provided es-

timates of total fat mass, lean body mass, and bone mass.

From these heights/lengths and weights, ponderal index in infancy (weight/length3) and BMI in child- hood was calculated (weight/height2). Body weight,

height/length, and BMI at each time point were con- verted to standard deviation (SD) scores (or z-scores)

by comparison with the British 1990 growth refer- ence106 using gestational age or the actual age at

measurement for each individual. Weight gain was as- sessed by calculating the difference in z-scores between

body weight at the beginning and end of the period, adjusting for regression toward the mean using LMS

correlates from the British 1990 growth reference.107

Fat mass index (FMI) was calculated by dividing fat

mass (in kg) by height (in mx) to adjust for body size.108 The optimal power (x) to raise height to was

derived from the data so that the relation between fat mass and height was completely removed108; this

power varied according to the age and sex of the child (e.g., 5.3 and 4.2 at age 11 y for boys and girls, respec-

tively109). There is no generally accepted cut-off to define excess adiposity using either fat mass, percent-

age of body fat, or FMI. However, if age- and sex- specific BMI cut-offs from the International Obesity

Task Force (IOTF)110 are used to assess the occurrence of overweight in subjects, a 20% prevalence of over-

weight is found. For comparability, it was assumed that an equivalent percentage of children should be de-

fined as having excess adiposity; e.g., those children in the top quintile of log FMI were categorized in

this way.

Blood samples and blood pressure

Nonfasting venous blood samples were collected at re- search clinics at some time points; fasting samples were collected only in a subgroup of the children at approxi-

mately 8 years. Various biochemical markers were mea- sured, and the methods are described in the individual

papers. Systolic and diastolic blood pressures were mea- sured at several ages using a Dinamap 9301 Vital Signs

Monitor (Morton Medical, London, UK). Two right arm measurements were recorded using a cuff size ap-

propriate for the child’s upper arm circumference, and the average was taken.

Pubertal status and girls’ age at menarche

Pubertal status was self-reported by questionnaire at

ages 11 years and 13 years using diagrams depicting the

Nutrition ReviewsVR Vol. 73(S3):175–206 181

5 Tanner Stages111 for pubic hair development and was

found to be reliable for assessing maturation in this co- hort. Data on whether or not the girls had started men-

struating were collected at a clinic held just before they reached 13 years of age; the mean (SD) age of atten-

dance at the clinic was 12.89 (0.23) years, ranging from 11.30 years to 14.34 years.43 The girls were asked during a measuring session in a private room if they had

started menstruating and, if so, when. In total, 3751 girls attended the clinic. Data were available from 3298

girls on whether or not they had reached menarche; of these, 1637 (50%) stated that they had not.

Maternal socioeconomic and anthropometric factors

Maternal age at delivery was calculated by subtracting

the mother’s date of birth from the child’s date of birth. Information on highest maternal educational level was

derived from a questionnaire sent out at 32 weeks gesta- tion. All of the educational qualifications obtained by

the mother were recorded, and her highest educational attainment was assessed on a 5-point scale: no academic

qualifications; vocational training (hairdressing, cater- ing, etc.); at least one O-level academic examination or

equivalent usually taken at age 16 years; at least one A-level academic examination or equivalent usually

taken at age 18 years; university degree. In some analy- ses, these categories have been further contracted to

only 3 categories; low, no academic qualification or vo- cational; medium, O-level qualifications; high, A-level

or degree-level qualifications. Information on maternal smoking status was collected by questionnaires sent

during pregnancy and at various ages of the child78. Housing tenure data were collected during pregnancy;

categories were as follows: owned or mortgaged; public or council rented; private rented. Mothers were asked

via questionnaire during pregnancy to report their pre- pregnancy weight and height78; these values were used

to calculate maternal prepregnancy BMI.

Statistical methods

Because this review features a large number of pub-

lished articles, it is not practical to list all of the statisti- cal methods used in the individual articles. Descriptive

statistics of the main variables, the likely biases, and the handling of missing data are provided individually in

most of the articles cited. Where regression analyses were carried out, the factors adjusted for depended on

the outcome of interest, and the variables used in each analysis are listed in the text, as appropriate. The main

confounding variables for many of the analyses were the indicators of SEB listed in Table 2. In all cases, 95%

CIs are quoted.

RESULTS AND COMMENTARY

Food and nutrient intakes throughout childhood

Longitudinal dietary change. The energy and energy-ad-

justed macronutrient intakes from the cross-sectional

food records at all 6 ages previously published in sepa-

rate articles are presented together in Table 4.15–18 For

the children aged 10 years and 13 years, among whom

misreporting rates were high, data from plausible repor-

ters only were used. The full nutrient details are avail-

able in the individual articles cited in Table 4. Energy

intake rose in line with the increasing size of the chil-

dren, as expected; within this, the contribution from

protein fell whereas that from carbohydrate rose. The

percentage of energy from protein was highest at age

1.5 years, then relatively constant from age 5 years on-

ward (decline of 1.2 [95% CI, 1.0%–1.4%] between age

1.5 years and age 3 years, n¼775; 0.7 [95% CI, 0.5%– 0.9%] between age 3 years and age 5 years, n¼636; both P < 0.001 by paired t-test). The increase in energy

from carbohydrate was mainly due to a rise in free sug-

ars; the percentage of energy from free sugars rose by

3.3 (95% CI, 2.9%–3.8%) between age 1.5 years and age

3.5 years and by 1.2 (95% CI, 0.7%–1.7%) between age 3

years and age 5 years (both P < 0.001). Fiber intake rela-

tive to EI was very similar at each age. Longitudinal change in EI was examined in com-

plex statistical analyses, combining all the food record and FFQ data collected on the children between ages 3

years and 13 years.19 Linear-spline multilevel models were used to summarize EI trajectories through child-

hood and adolescence. Models estimated average and individual EI at age 3 years and linear changes in EI

from age 3 years to age 7 years and age 7 years to age 13 years in 12 032 children with at least 1 food record or

FFQ and were adjusted for sex, person completing food record (parent, child), source of data (FFQ, food re-

cord), and plausibility of intake. Trajectories were simi- lar if FFQs and food records were analyzed separately

or combined into one model. The FFQ estimation of EI was, on average, 74 kcal higher than that of the food re-

cords. The predicted EI at exactly 3 years was 1131 (SD, 74) kcal. The rate of increase of EI was greater between

ages 3 years and 7 years (136 [SD, 16] kcal/y) than be- tween ages 7 years and 13 years (97 [SD, 20] kcal/y).

The percentage of consumers and overall amount of selected food groups from separate publications at each

age as the children progressed from preschool age to ad- olescence are shown in Table 5.15–18 Visual inspection of

the data gives an insight into the relative quality of the diet over time. The data suggest that the profile of foods

consumed moved away from the food-based recommen- dations of the UK Eatwell plate100 and the Australian

182 Nutrition ReviewsVR Vol. 73(S3):175–206

core and noncore food concept.94 Within the core foods,

those high in fiber (wholemeal/brown bread and whole- grain breakfast cereals) showed a relative decline in per-

centage of consumers, whereas white bread and other breakfast cereals showed an increase, at least up to age 7

years. Beef products were consumed by many fewer chil- dren at age 3.5 years (25%) than age 1.5 years (37%),

probably a result of major publicity about bovine spongi- form encephalitis in beef, which occurred just prior to

collecting food records for the children aged 3.5 years; these data suggest that consumption took several years to

recover back to the earlier frequency (35% by age 10 years). Beef was eaten by 45% of children aged 1.5–3

years in NDNS 2008–2012,102 confirming the recovery of intake. More children ate oily fish as they grew up, but

never more than 20% of the ALSPAC children consumed it. More than 10% of children had not eaten any vegeta-

bles, and more than 15% had not eaten any fruit during the recording period at each age (Table 5). The data sug-

gest that whole milk was gradually replaced by low-fat milk in the diets of these children. In the noncore food

groups, savory snacks (crisps), biscuits, and chocolate confectionery were consumed by a majority of the chil-

dren at each age, and the percentage of consumers of

sugar confectionery more than doubled after age 1.5

years. Coated poultry products were consumed most of- ten at age 7 years; they were widely used in school din-

ners at this time prior to reforms in school meal quality. Regular and/or diet soft drinks were consumed by a ma-

jority of the children at each age in relatively large amounts (Table 5).

A detailed picture of the changes in the sources of energy from the food groups at the point of greatest

change in the diet between age 1.5 years and age 3.5 years are given in Table 6.23 The decline in energy con-

tribution from dairy foods (P < 0.001) probably ac- counted for the decreased percentage of energy from

protein (Table 4), whereas the free sugars increase was likely due to the increased EI from sweet miscellaneous

foods (P < 0.001), particularly confectionery. The con- tribution to EI of core foods fell from 63% to 54%,

whereas the contribution of noncore foods rose from 32% to 43% (both P < 0.001). Of the 11% rise in energy

from noncore foods, 5% was due to increased intakes of confectionery and savory snacks (crisps). These changes

suggest deterioration in diet quality between the 2 ages, and Table 5 suggests that this lower-quality diet is

maintained throughout childhood.

Table 4 Mean intake of energy and fiber and the percentage of energy from macronutrients at each age of food record assessment from age 1.5 years to age 13 years Item measured Age at assessment P-value Age at assessment P-value Age at assessment P-value

1.5 ya 3.5 yb (1.5 y vs 3.5 y)

5 yc 7.5 yd (5 y vs 7.5 y)

10.5 ye 13 yc (10.5 y vs 13 y)

(plausible only)

(plausible only)

No. of respondents 1026 863 772 814 4462 2227 Plausible reporters (%) 73 69 72 76 60 35 Energy (MJ) 4.60 [0.94] 5.65 [1.1] 6.37 [1.23] 7.06 [1.32] 8.39 [1.21] 9.90 [1.59] Protein (%en) 15.3 [2.3] 14.0 [2.5] <0.001 13.4 [2.3] 13.3 [2.2] 0.998 13.3 [2.3] 13.7 [2.5] <0.001

(15.1–15.4) (13.9–14.2) (13.2–13.5) (13.2–13.3) (13.2–13.4) (13.6–13.8) Total fat (%en) 37.4 [4.7] 36.2 [4.5] <0.001 35.5 [4.5] 35.5 [4.4] 0.838 35.9 [4.6] 35.6 [5.2] 0.409

(37.1–37.7) (35.9–36.5) (35.2–35.8) (35.4–35.6) (35.8–36.0) (35.4–35.9) Saturated fat (%en) 18.0 [3.6] 16.1 [3.3] <0.001 14.8 [3.0] 14.2 [2.8] 0.028 14.0 [2.9] 13.6 [3.2] <0.001

(17.7–18.2) (15.9–16.3) (14.6–15.0) (14.2–14.3) (13.9–14.0) (13.5–13.7) Monounsaturated

fat (%en) 12.0 [1.8] 11.8 [1.7] 11.9 [1.8] 11.9 [1.9] 12.2 [2.0] 11.9 [2.3]

(11.9–12.1) (11.7–12.0) (11.8–12.0) (11.9–12.0) (12.2–12.3) (11.8–12.0) Polyunsaturated

fat (%en) 4.3 [1.6] 5.0 [1.6] 5.4 [1.7] 5.5 [1.6] 5.9 [1.8] 5.8 [1.9]

(4.2–4.3) (4.9–5.1) (5.3–5.5) (5.5–5.6) (5.8–5.9) (5.7–5.9) Carbohydrate (%en) 46.7 [6.0] 49.9 [5.0] <0.001 51.2 [5.4] 51.1 [5.1] 0.944 50.8 [4.9] 50.6 [5.9] 0.037

(47.1–47.7) (49.5–50.2) (50.8–51.5) (51.2–51.4) (50.7–51.0) (50.4–50.9) Starch (%en) 20.8 [5.2] 23.8 [5.5] 25.4 [4.7] 26.5 [4.4] 26.6 [4.4] 26.8 [5.1]

(20.5–21.1) (23.4–24.1) (25.1–25.7) (26.4–26.6) (26.5–26.7) (26.6–27.0) Free sugars (%en) 12.3 [6.0] 16.4 [5.9] <0.001 17.1 [6.0] 17.4 [5.3] 0.002 18.1 [5.5] 17.9 [6.5] 0.021

(12.0–12.6) (15.5–16.3) (16.7–17.5) (17.3–17.5) (17.9–18.3) (17.6–18.1) Fiber (g of nonstarch

polysaccharide) 6.6 [2.4] 8.1 [2.8] 9.3 [3.1] 10.3 [3.3] 12.1 [3.5] 14.6 [4.8]

(6.5–6.8) (7.9–8.3) (9.1–9.5) (10.1–10.5) Data are presented as mean (SD) and (95% confidence intervals). These data were first published in the articles cited; for compari- sons in this table, paired t-tests have been undertaken between certain groups. a Data from Cowin et al. (2000)15 b Data from Emmett et al. (2002)16 c Not published in this format previously. d Data from Glynn et al. (2005)17 e Data from Cribb et al. (2011)18 Abbreviations: %en, percentage of energy.

Nutrition ReviewsVR Vol. 73(S3):175–206 183

Comparisons with dietary recommendations and cross-

sectional dietary intake data from representative samples of UK children. The profile of fat intake improved

slightly over childhood (Table 4), with a fall in the over- all contribution of fat to energy toward the recom-

mended level of <35%.15–18 This was mainly due to a

fall in energy from saturated fat, although intake was

still above the recommendation of 10% of energy. The amount of free sugars consumed at all ages was much

higher (Table 4) than the maximum 10% of energy rec- ommended.93 Fiber (nonstarch polysaccharide) intakes were approximately 75% of the recommended adequate

Table 5 Percentage of children who consumed selected core and noncore food/beverage groups and the daily mean (SD) weight consumed (averaged over consumers and nonconsumers) at each age of food record assessment Item consumed Age at assessment

1.5 ya 3.5 yb 5 yc 7.5 yd 10.5 ye 13c

(plausible only) (plausible only) No. of respondents 1026 863 772 814 4462 2227 Core foods and beveragesf

White bread (%C) 73.7 83.6 85.1 89.4 88.0 80.5 White bread (g) 20.3 (21.0) 35.6 (29.5) 48.4 (38.1) 54.3 (38.2) 64.6 (47.4) 68.4 (62.1) Wholemeal/brown bread (%C) 34.3 32.0 26.6 22.8 24.5 21.5 Wholemeal/brown bread (g) 7.9 (15.0) 10.6 (19.6) 11.1 (23.8) 9.2 (21.6) 7.6 (21.6) 11.9 (30.7) High fiber breakfast cereal (%C) 72.7 59.0 57.1 49.8 49.0 48.1 High fiber breakfast cereal (g) 13.9 (18.2) 13.7 (20.5) 14.1 (19.7) 15.3 (21.7) 15.6 (24.8) 21.9 (35.1) Other breakfast cereal (%C) 45.8 65.7 63.7 57.0 51.5 38.7 Other breakfast cereal (g) 4.5 (7.5) 9.7 (10.8) 12.5 (14.9) 14.2 (17.0) 13.4 (17.8) 12.7 (21.3) Beef (%C) 36.8 24.9 26.1 30.0 35.4 36.2 Beef (g) 9.2 (18.0) 10.3 (25.4) 8.9 (18.2) 11.7 (22.8) 21.7 (41.0) 29.0 (55.2) Oily fish (%C) 11.4 13.6 13.5 15.4 16.5 19.7 Oily fish (g) 1.4 (5.7) 2.6 (8.5) 3.0 (9.1) 3.5 (10.1) 4.8 (14.3) 7.8 (23.9) All vegetables (%C)g 91.6 83.0 83.3 88.1 85.2 86.3 All vegetables (g)g 39.8 (31.7) 40.5 (36.7) 44.6 (40.4) 50.7 (43.4) 65.4 (56.3) 80.1 (69.2) All fruit (%C) 84.1 82.6 82.1 77.8 64.0 72.1 All fruit (g) 63.9 (56.5) 68.9 (63.6) 78.8 (70.8) 75.7 (74.5) 73.3 (77.8) 92.8 (106.1) Plain potatoes (%C) 78.3 62.6 57.3 56.1 51.5 48.6 Plain potatoes (g) 29.2 (27.1) 25.4 (30.1) 25.0 (30.1) 26.8 (32.0) 34.9 (46.0) 40.8 (57.5) Whole milk (%C) 90.2 83.7 66.6 55.2 34.5 25.1 Whole milk (g) 391 (243.9) 260.9 (231.8) 173.6 (189.8) 138.2 (189.8) 80.3 (152.0) 69.9 (174.2) Semi/skimmed milk (%C) 14.6 33.3 44.3 54.0 61.5 65.5 Semi/skimmed milk (g) 30.4 (113.7) 76.5 (159.2) 94.0 (151.6) 122.6 (168.1) 144.7 (179.6) 178.3 (220.4)

Noncore foods and beveragesf

Coated poultry (%C) 10.4 24.1 29.8 34.8 27.0 16.1 Coated poultry (g) 1.6 (5.4) 5.0 (11.3) 8.3 (15.3) 10.7 (17.9) 10.4 (22.2) 8.9 (26.4) Potatoes with fat (%C) 56.1 72.3 77.2 83.2 84.0 68.8 Potatoes with fat (g) 17.6 (23.7) 32.4 (34.1) 39.2 (34.0) 43.4 (38.1) 63.0 (54.3) 64.9 (66.4) Biscuits (%C) 87.2 87.7 83.5 82.9 79.5 72.7 Biscuits (g) 11.0 (10.4) 15.6 (13.4) 19.1 (17.4) 19.9 (17.2) 21.6 (21.6) 24.9 (28.8) Cakes, buns (%C) 46.6 55.4 62.3 68.6 70.0 63.8 Cakes, buns (g) 8.1 (12.4) 16.0 (21.5) 23.3 (27.6) 25.7 (27.2) 31.1 (33.4) 38.0 (44.8) Puddings and ice cream (%C) 57.9 69.2 70.7 70.3 68.5 51.8 Puddings and ice cream (g) 24.9 (33.9) 35.8 (39.6) 39.6 (41.1) 43.8 (46.1) 48.4 (53.0) 37.9 (54.9) Savory snacks (%C) 65.6 75.2 78.9 83.2 82.0 72.5 Savory snacks (g) 5.9 (7.2) 11.2 (10.9) 15.0 (12.1) 17.2 (12.9) 20.3 (16.3) 18.2 (18.0) Chocolate confectionery (%C) 63.3 67.8 67.7 74.8 77.5 68.0 Chocolate confectionery (g) 6.9 (8.6) 11.6 (13.1) 11.4 (12.9) 16.3 (17.3) 20.8 (22.2) 22.8 (27.6) Sugar confectionery (%C) 18.4 41.9 45.5 43.9 53.5 38.8 Sugar confectionery (g) 1.8 (5.3) 5.3 (9.4) 6.2 (10.7) 6.1 (10.6) 9.1 (16.1) 10.6 (25.9) Normal soft drinks (%C) 40.3 59.0 66.7 62.3 62.4 59.5 Normal soft drinks (g) 73.4 (190.3) 116.9 (198.1) 157.2 (232.7) 147.1 (208.5) 158.5 (211.4) 215.8 (304.2) Diet soft drinks (%C) 65.4 75.0 72.7 73.1 69.4 56.3 Diet soft drinks (g) 172.5 (248.0) 286.9 (344.7) 285.1 (345.2) 273.7 (330.2) 227.2 (263.2) 249.0 (370.1)

a Data from Cowin et al. (2000)15 b Data from Emmett et al. (2002)16 c Not published in this format previously. d Data from Glynn et al. (2005)17 e Data from Cribb et al. (2011)18 f Data have not been compared by formal statistics. g Excludes baked beans. Abbreviation: %C, percentage of children who consumed indicated food or drink; SD, standard deviation.

184 Nutrition ReviewsVR Vol. 73(S3):175–206

intake at each age throughout childhood (Table 4); e.g.,

an adequate intake for children aged 3 years is approxi- mately 10.8 g nonstarch polysaccharide/day, but the

mean intake in these children at age 3.5 years was 8.2 g/ day.16,99

Most micronutrient intakes were adequate at all ages when compared with the recommendations,93 the exceptions being low intakes of vitamin D and iron.

Average intakes for iron were below the recommended amounts at all ages in childhood.15–18 Current recom-

mendations for Vitamin D cover only children aged 3 years or below.

Intakes of both foods and nutrients were similar to those in comparable age groups in NDNS 1992/1993,101

NDNS 1997,92 and NDNS 2008–2012.102 Full compari- sons are presented in each of the ALSPAC publications

covering childhood diet.15–18 In the most recent survey, the children aged 1.5–3 years in NDNS 2008–2012 con-

sumed more fruit than the ALSPAC children; 93% ate fruit in 2008–2012 compared with approximately 83%

in ALSPAC in 1993–1996. This was due to fruit being

provided to children in school breaks free of charge. In

NDNS 2008–2012, free sugars contributed 11.9% of the energy in the youngest group, increasing to 14.8% and

15.8% at ages 4–10 years and 11–18 years, respec- tively,102 paralleling the ALSPAC findings.

Furthermore, 31% of children aged 1.5–3 years in the NDNS ate sugar confectionery, increasing to 49% of children aged 4–10 years,102 which is very similar to the

ALSPAC levels (see Table 5). This suggests that high sugar consumption is still prevalent in children.

Dietary differences in relation to family socioeconomic

background. The quality of the diet in relation to family SEB was investigated in preschool children at ages 1.5

years and 3 years.20 The strongest associations for diet were with maternal education: the nutrient differences

found were lower intakes of nonstarch polysaccharide and many micronutrients and higher intakes of free

sugars in children of the mothers with low compared with high education. Differences in foods consumed

were split mostly along core and noncore food lines:

Table 6 Contribution to energy intake of core and noncore food groups from 3-day food records of diet consumed by the same children at age 1.5 years and age 3.5 years from the 10% subsample of the ALSPAC cohort (n¼755) Type of foods Energy supplied by

each food group (% of total)a

Type of foods Energy supplied by each food group (% of total)a

At age 1.5 y At age 3.5 y At age 1.5 y At age 3.5 y Core foods Noncore foods Bread, cereals, rice, and pasta Miscellaneous

Bread 7.1 9.2 Puddings and ice-creams 3.3 4.3 Breakfast cereals 5.6 5.6 Buns, cakes, and pastries 2.8 4.3 Pasta, rice and savories 2.7 3.4 Sweet biscuits 4.9 5.7 Group total 15.3b 18.2c Savory biscuits 0.7 0.6

Vegetables and legumes Confectionery 4.8 7.1 Vegetables 1.5 1.3 Crisps 2.5 4.3 Potatoes 2.1 1.7 Group total 19.1b 26.3c

Vegetable dishes 0.3 0.3 Processed meat, fish, and poultry Legumes 0.1 0.1 Processed meats 0.8 1.0 Group total 4.0b 3.3c Coated chicken 0.3 0.9

Fruit Burgers and kebabs 0.2 0.2 Fruit 4.4 3.4 Sausages 1.1 1.4 Fruit juice 1.5 1.8 Meat pies 0.8 0.9 Group total 5.9b 5.2c Coated and fried fish 1.0 1.3

Yogurt, cheese, and milk Group total 4.3b 5.8c

Milk 25.1 15.6 Vegetables Yogurt 4.0 3.2 Fried/roast potatoes 2.9 4.3 Cheese 2.3 2.6 Baked beans 1.2 0.9 Group total 31.4b 21.4c Group total 4.1b 5.3c

Spreads, soup, and sauces Meat, fish, poultry, and eggs Fat spreads 3.3 4.3

Meat 2.7 2.4 Soup 0.3 0.3 Fish 0.5 0.5 Milk-based sauces 0.2 0.1 Poultry 1.2 1.6 Tomato-based sauces 0.1 0.2 Eggs and egg dishes 1.2 1.1 Other sauces 0.5 0.6 Group total 5.7 5.6 Group total 4.3b 5.5c

Core foods total 62.8b 53.8c Noncore foods total 31.8b 42.8c

Modified from Cribb VL et al. (2013)23 with permission. a The percentage of energy from core and noncore foods does not add up to 100% because a few energy-containing foods have not been included in the classification. b,c Groups with different letters show a significantly different energy contribution between ages by paired t-test (P < 0.001).

Nutrition ReviewsVR Vol. 73(S3):175–206 185

cheese, yogurt, fruit, and fruit juice were consumed less

often and chocolate, savory snacks, meat products, and fried potatoes were consumed more often by children

of mothers with low compared with high education (all P < 0.001).

At age 10 years, child’s dietary differences relating to maternal education (low [n¼832], medium [n¼1472], and high [n¼1820]) were examined again.18 There was no difference in overall EIs between the education groups; however, energy-adjusted fat in-

take was slightly higher (by 3.1%) and carbohydrate and protein intake slightly lower (both by �2%) and fiber intake lower (by 6.1%) in low compared with high ma- ternal education groups (all P < 0.001). Intake of free

sugars was high in all groups (�85 g/d, which is twice the recommended maximum). The differences in some

micronutrients were substantial, e.g., vitamin C was 23% lower, carotene was 15% lower, retinol was 12%

lower, and sodium was 3% higher in the low compared with the high maternal education groups. Food group

differences were also evident, with the core foods more likely to be consumed by children of mothers with high

educational attainment; in the high maternal education group vs the low maternal education group, wholemeal

bread was consumed by 30.1% vs 18.5%, oily fish was consumed by 22.3% vs 10.6%, cooked vegetables were

consumed by 82.7% vs 73.6%, and fresh fruit was con- sumed by 81.7% vs 59.9%, respectively (all P < 0.001).

Conversely, noncore foods were consumed by a higher proportion of the low maternal education group com-

pared with the high maternal education group; meat pies and pasties were consumed by 20.9% vs 14.7%,

fried potatoes and chips were consumed by 84.9% vs 73.2%, crisps/savory snacks were consumed by 84.5% vs

78.0%, and diet drinks were consumed by 76.0% vs 61.0%, respectively (all P < 0.001).18 The overall

trend in both nutrients and foods was toward a more energy-dense, nutrient-poor diet as the educational at-

tainment of the child’s mother decreased.

Individual nutrients. Dietary fat. Recommendations

about the amount of dietary fat that is suitable for pre- school children are confusing. An investigation was

made into differences in nutrient, food group intakes, and growth between children divided into quartile

groups according to their intake of energy from fat.21

The average contribution to energy from fat across the

quartiles rose from 31.2% (SD, 2.8) in the lowest quar- tile to 43.1% (SD, 2.2) in the highest quartile at age 1.5

years and from 30.4% (SD, 2.5) to 41.8% (SD, 2.0), re- spectively, at age 3.5 years. There was an increase in EI

between the lowest and the highest fat intake quartiles (by 7.4% at age 1.5 years and 5.5% at age 3.5 years; both

P < 0.001). However, there was a drop in total

carbohydrate intake between the lowest and the highest

fat intake quartiles, which was driven by a drop in in- take of free sugars (by 44% at age 1.5 years and 29% at

age 3.5 years; both P < 0.001). Some vitamins and min- erals increased with increasing fat intake quartile, par-

ticularly retinol equivalents and zinc; however, intakes of iron and most water-soluble vitamins fell. Vitamin C intake was 49% and 39% lower in the highest than the

lowest quartile of fat intake at ages 1.5 years and 3.5 years, respectively, whereas the equivalent differences

for iron intake were 23% and 19% lower (all P < 0.001). The foods increasing with rising fat intake were whole

milk, meat products, cheese, and crisps/savory snacks (all P < 0.001). Those decreasing with increasing fat in-

take were semi-skimmed milk, fruit, fruit juice, break- fast cereal, and fish (all P < 0.001). There were no

differences in growth among children in the 4 dietary fat quartiles; height and BMI at age 1.5 years and age

2.5 years were not different between quartiles at age 1.5 years, and height and BMI at age 3.5 years and age 5.0

years were not different between quartiles at age 3.5 years.22 There was also no difference in the proportion

of children in each quartile that were at either extreme of height or BMI at any of the ages investigated. These

results suggest that within the range of fat intakes nor- mally consumed by United Kingdom preschool chil-

dren, there is no evidence of an effect on growth or obesity development of increasing the proportion of EI

coming from fat in the diet.

Vitamin A. The sources of vitamin A (retinol equivalents) in the diet and changes in intake between age 1.5 years

and age 3.5 years were examined longitudinally (n¼755 with food records at both ages).23 The diet was categorized by core and noncore foods, and there was evidence that

the quality of the diet deteriorated with age.23 Between age 1.5 years and age 3.5 years, the dietary intake of vitamin A

decreased by 56 mg/day (95% CI, �17 to �95 mg/d; P¼0.005) mainly due to decreased intake of whole milk (390 g/d at age 1.5 y; 261 g/d at age 3.5 y) and increased in- take of energy-dense, nutrient-poor foods such as pud-

dings, cakes, biscuits, confectionery, and crisps. The contribution to energy of these foods rose from 19.1% to

26.3% (P < 0.001), with very little contribution to vitamin A intake. These analyses highlight the importance of feed-

ing young children foods that are rich in nutrients in place of energy-dense, nutrient-poor foods.

Vitamin D. A parallel analysis investigated vitamin D and calcium intakes and changes with age between

1.5 and 3.5 years.24 Overall, vitamin D intakes were low; all children had intakes below the UK dietary recom-

mendations at both ages. Calcium intakes decreased between the 2 ages due to reduced milk/dairy

186 Nutrition ReviewsVR Vol. 73(S3):175–206

consumption. Children in the lowest quartile for vita-

min D intake at 1.5 years were twice as likely to remain in that quartile at 3.5 years (odds ratio [OR], 2.35; 95%

CI, 1.56–3.55). Only 18% of children consumed supple- ments containing vitamin D at age 1.5 years, falling to

11% at age 3.5 years. In NDNS 2008–2012, young chil- dren obtained only a quarter of the vitamin D intake recommended to be obtained from diet and only 12%

took dietary supplements102; thus, there is a continuing problem in the United Kingdom. Whether fortification

of foods could be the answer to this problem was exam- ined; the theoretical intakes from different fortification

regimens tested suggested that milk fortified at 2 mg vi- tamin D/100 g would provide most preschool children

with adequate but not excessive intakes.24 Later work in the ALSPAC looking at plasma concentrations in mid-

childhood of vitamin D (25-hydroxyvitamin-D2 and 25-hydroxyvitamin-D3) in relation to cortical bone

measures found positive associations with D3 only, sug- gesting that in supplementation or fortification, D3 should be used in preference to D2.

25 These 2 papers give valuable insights into how the lack of vitamin D in

the diets of the majority of UK residents could be tackled.

Iron. Iron is a nutrient that is essential to a child’s

growth and development. Several of the ALSPAC arti- cles reporting on diet in childhood identify that a size-

able proportion of children, particularly girls,15–18,22

have iron intakes that are below recommended

amounts throughout childhood. This is especially wor- rying for girls because their iron needs increase as

they enter adolescence due to increased losses from menstruation.

Dietary energy density. The food energy density of the diet was investigated in the CIF subsample at age 5

years and age 7 years with the energy contribution from drinks considered as a separate variable in the analy- sis.26 Overall, the energy density of the diet was approx-

imately 8.5 at age 5 years and approximately 8.8 kJ/g at age 7 years, and there was relatively strong tracking be-

tween the 2 ages (intraclass correlation coefficient: 0.62; 95% CI, 0.55–0.68), suggesting that children with a

highly energy-dense diet at age 5 years were likely to have a diet with a high energy density at age 7 years.

There was no evidence of an association between die- tary energy density and EI at age 5 years, but at age 7

years there was a weak association (r¼0.15), suggesting that diets with a high energy density at 7 years have a

slightly increased energy content. At both ages, energy- dense diets have higher fat and lower fiber content than

diets that have a low energy density.26

The energy density of the diets at age 10 years was

calculated (without drinks) for another study.27 Mean energy density was 8.76 (SD, 1.63) kJ/g, and at this age,

increasing energy density was strongly associated with increasing overall EI (r¼0.22; P < 0.0001). Energy den- sity was lower among under-reporters than plausible re- porters (8.45 [SD, 1.67] kJ/g vs 8.87 [SD, 1.56] kJ/g; P < 0.0001), underlining the importance of accounting

for reporting status in any analyses using these data. The data suggest that the positive relationship between

energy density and EI becomes stronger as children grow up.

Individual food groups. Fruit and vegetable intakes. Determinants of fruit and vegetable intakes were inves-

tigated using the whole ALSPAC cohort of children aged 7 years who provided food records (n¼7285).28 The possible determinants assessed were mother’s con- sumption of fruit and vegetables (from an FFQ com-

pleted when the child was aged 4 y), provision of fruit and vegetables at meals in the home, child’s eating be-

haviors (liking to try a variety of foods, choosiness, and enjoyment of food), maternal education, and family ex-

penditure on food per head.28 Total fruit consumption in boys and girls was 123 and 133 g/day, and total vege-

table consumption was 71 and 72 g/day, respectively, confirming that boys consumed less of fruit than girls.

However, there was no difference between boys and girls in the determinants of intake, so they were com-

bined for this analysis. For children to achieve the rec- ommended 5 portions of fruit and vegetables per day,

total intake should be around 45 g/MJ/day (equating to 315 g/d for a 7-year-old child). The average intake in

these children was just above half of this amount. For fruit intake (on average, 18 g/MJ/d), in a model

that only included items that were independently re- lated to fruit intake (n¼5259), 14% of the variance in intake was explained.28 Providing fruit as part of every- day meals compared with not doing so had the stron- gest independent association (adjusted difference, 7.2;

95% CI, 6.9–8.3 g/MJ/d higher intake; P < 0.001); there were also associations for the highest tertile of maternal

fruit consumption compared with the lowest (adjusted difference 6.3; 95% CI, 5.3–7.3 g/MJ/d; P < 0.001), high

maternal educational attainment compared with low (adjusted difference 5.7; 95% CI, 4.6–6.8 g/MJ/d;

P < 0.001), the highest level of expenditure on food compared with the lowest (adjusted difference 3.8; 95%

CI, 2.4–5.0 g/MJ/d; P < 0.001), and the child not being choosy about food compared with being very choosy

(adjusted difference 2.8; 95% CI, 1.6–3.9 g/MJ/d; P < 0.001). The child liking variety or enjoying food

was not associated with fruit intake.

Nutrition ReviewsVR Vol. 73(S3):175–206 187

For vegetable intake (on average, 10 g/MJ/d), in a

similar model (n¼5208), child factors were much more important than they had been for fruit intake, all vari-

ables explained 9% of the variance.28 The child not being choosy compared with the child being very choosy was

associated with the highest difference (adjusted differ- ence, 2.6: 95% CI, 1.9–3.3 g/MJ/d higher intake; P < 0.001); the child liking compared with not liking to

eat a variety of foods also had a strong association (ad- justed difference, 1.5; 95% CI, 0.9–1.9 g/MJ/d; P < 0.001).

Maternal factors included regularly providing vegetables at family meals compared with not doing this (adjusted

difference, 1.7; 95% CI, 1.3–2.2 g/MJ/d; P < 0.001), mother being in the highest tertile of vegetable intake

compared with the lowest (adjusted difference, 1.7; 95% CI, 1.2–2.1 g/MJ/d; P < 0.001), and mother having a high

compared with a low educational attainment (adjusted difference, 1.2; 95% CI, 0.7–1.7 g/MJ/d; P < 0.001).

Expenditure on food and child enjoyment of food were only weakly associated (P < 0.05).28 In summary, mater-

nal example and regular provision are important in de- termining children’s intakes of fruits and vegetables, and

a child’s eating behavior adds another dimension, partic- ularly in relation to vegetable eating.

Drinks in the diet. The type and volume of drinks con-

sumed was investigated in detail in the CIF subsample at age 5 years and age 7 years.29 The analysis focused on

sugar-sweetened beverages but assessed other drinks as well. At both ages, the most popular drink was milk (me-

dian intake, 257 and 242 g/d at age 5 y and age 7 y, re- spectively); the next most popular drinks were artificially

sweetened diet soft drinks (127 and 140 g/d), with water (56 and 75 g/d) and sugar-sweetened beverages (57 and 67 g/d) at similar amounts. There was a bias in consump-

tion of some of the drinks by maternal education; at both ages, children with degree-educated mothers con-

sumed the largest amounts of fruit juice and water and the smallest amounts of diet soft drinks; children whose

mothers had vocational training consumed the largest amount of diet drinks (all P < 0.01).29 Drinking diet

soft drinks may disrupt appetite control mechanisms because sweetness is disassociated from energy con-

tent.112 Child intakes of sugar-sweetened beverages and milk did not differ according to maternal education

groups.

Physical activity and diet. Objectively measured physical

activity was available at age 11 years; this was combined with dietary intake at age 10 years to test whether there

was a demonstrable relationship between diet and phys- ical activity (n¼5134 with diet and physical activity; n¼3684 if restricting to plausible reporters only).30

Physical activity was higher in boys than in girls (in the

full sample; means of 664 and 552 counts per minute, respectively). The amount of time spent in moderate to

vigorous physical activity was also higher in boys than in girls (mean of 28 min compared with 18 min/d on a

weekday; all P > 0.005). Boys recorded higher energy consumption than girls (mean of 1952 compared with 1769 kcal/d); however, boys consumed a lower weight

of fruit and vegetables than girls (mean of 136 com- pared with 147 g/d).30 Only weak associations between

physical activity and dietary intake were found, which differed between boys and girls: the most consistent as-

sociations were in boys, with total energy and percent- age of energy from carbohydrate positively related

(both P¼0.007) and percentage of energy from fat neg- atively related (P¼0.021); the only association in girls was with fruit and vegetable consumption, which was positively related (P¼0.027). Using plausible reporters only did not change the associations greatly. It is possi- ble that associations between diet and physical activity

would have been stronger if diet had been measured at the same time as physical activity; however, these data

suggest that the 2 behaviors are relatively independent of each other.

Plausibility of dietary energy reporting. The plausibility of the dietary EIs recorded at each age was investigated

in relation to predicted energy requirements, and in gen- eral, the frequency of likely under-reporting of EI in-

creased and the frequency of likely over-reporting of EI decreased with age.15–18 The frequency of misreporting

was compared in the same children at age 5 years and age 7 years (plausible reporters: 72% and 76%, respec-

tively).26 There was evidence of tracking between these ages (intraclass correlation coefficient: 0.41: 95% CI,

0.34–0.48), suggesting that those who under-reported EI at age 5 years were slightly more likely to under-report at

age 7 years as well.26 In the food records collected at age 10 years18 and age 13 years, the frequency of under-re- porting was high (36% and 62%, respectively) but in line

with NDNS 199792 frequency for similarly aged children. At age 10 years, children who were overweight or obese

were more likely than those of normal weight to under- report their intake (P < 0.001).18 There were differences

in numbers of consumers and average intakes for partic- ular foods, with under-reporters recording lower intakes

than plausible reporters of some core foods (rice and pasta, whole milk; in girls only: cooked vegetables; in

boys only: fresh fruit [all P < 0.001]) and some noncore foods (biscuits, cakes, puddings, chocolates, sweets, sweet

spreads, and sugar-sweetened beverages in both sexes [all P < 0.001]). These data suggest that particular types of

noncore, energy-dense foods tend to be under-reported.

188 Nutrition ReviewsVR Vol. 73(S3):175–206

A further complication in the identification of mis-

reporting of dietary intakes is the possibility that differ- ences in physical activity between individuals are

distorting associations between diet and biological out- comes.31 At age 13 years, assessment of both diet and

physical activity with reasonable objectivity was avail- able; therefore, an investigation was made of whether the inclusion of measured physical activity in equations

used to assess diet plausibility would provide clearer re- lationships.31 Three methods for predicting energy re-

quirements were used: one allowed for standard low physical activity, the second calculated an individual’s

physical activity from prediction equations, and the third used measured minutes of MPVA; all methods in-

cluded an allowance for the age, sex, and size of the child. The frequency of plausible reporting was very

similar among the methods (�40%), but the use of measured physical activity gave much lower estimates

of under-reporting (37.1%) and higher estimates of over-reporting (20.4%) compared with the first 2 meth-

ods (under-reporting 51.5% and 51.8%; over-reporting 7.7% and 10.3%, respectively). For all 3 methods, un-

der-reporters had higher mean BMI and waist circum- ference than plausible or over-reporters, as well as a

higher percentage body fat (all P < 0.001). Conversely, percentage lean body mass was lowest in under-

reporters and highest in over-reporters (all P < 0.001).31

Because these 13-year-old children were mostly respon-

sible for reporting their own diet with minimal parental help, it is very likely that they missed some foods; there-

fore, high frequency of under-reporting would be ex- pected. The fact that using the measured physical

activity to assess misreporting resulted in a shift toward over-reporting suggests that total physical activity was

underestimated with this method. Further work should find ways of incorporating assessment of light physical

activity, as well as moderate to vigorous physical activ- ity, into this method to improve estimates of total physi-

cal activity.

Achieving healthy diet recommendations. The dietary

data collected at 1.5 and 3.5 years was used to inform a study that aimed to provide appropriate portion size

ranges and a practical food plan that covered all food and nutrient recommendations for use with children

aged 1–4-years.32 Food and portion size information from the NDNS was used80 in combination with the

ALSPAC data. To design a food plan to provide an ade- quate nutrient content within the recommended energy

requirements for children, it was necessary to use foods with a medium or high nutrient density. The inclusion

of energy-dense foods with poor nutrient content led to unbalanced nutrient-to-energy content in the food

plan, so these foods had to be kept to a minimum. It

proved impossible to incorporate enough vitamin D–

containing foods to cover the recommendations, imply- ing that vitamin D–deficient diets are likely to be the

norm.32

In the mid-2000s, there was a great deal of contro-

versy about the nutritional content of school meals. Children with food records for either packed lunches (n¼410) or school dinners (n¼211) from the CIF sub-study of 7-year-olds were identified,33 and the qual- ity of school meals as consumed was examined. In gen-

eral, packed lunches had a less good nutrient profile than school dinners, with their saturated fat and sugar

content being higher and their micronutrient content being lower. Both types of meal were inadequate when

compared with nutrient guidelines.93 School dinners in the United Kingdom have been improved since these

meals were recorded in 2000. The packed lunches re- corded were compared with food-based guidelines of

what a school meal should contain99: 1 item from each core food group (starchy foods; fruits; vegetables; milk

or dairy; meat, fish, or protein alternative). Only 3.5% contained all 5 groups; 44.3% contained <2 of the

groups; 31% did not include a protein food.33 There was a fruit in only 41% and a vegetable/salad item in

only 16% of the packed lunches. Taken as part of the whole days’ intake, the packed lunch eaters had a

slightly higher overall intake of energy-adjusted satu- rated fat and sugar, even after controlling for maternal

education (both P¼0.014); they also had a lower potas- sium intake but higher selenium intake (both P < 0.001)

than school dinner eaters. This work highlights the fact that provision of nutritionally adequate cooked school

dinners is needed and suggests that this should be ac- companied by an education campaign for parents re-

garding what constitutes a nutritious packed lunch.33

Growth and obesity development. Sensitivity of assess-

ment methods. When assessing growth and obesity de- velopment in childhood, it is necessary to standardize weight, height, and BMI against growth reference

curves. Most analyses using ALSPAC growth data have used the UK 1990 growth references;106 however, new

growth standards for children up to age 5 years were published by the World Health Organization in 2006,113

and the effect of using these in place of the UK refer- ence for growth monitoring was investigated.34 Using

the World Health Organization 2006 standards resulted in fewer ALSPAC infants being classified as under-

weight in the first year and more ALSPAC preschool children being classified as overweight (at age 2 y OR

for overweight, 1.74; 95% CI, 1.20–2.51 and at age 5 y OR, 1.35; 95% CI, 1.02–1.78). It is essential to specify

which growth reference has been used in any study

Nutrition ReviewsVR Vol. 73(S3):175–206 189

investigating growth and obesity development in

children. Two methods of identifying obese children using

BMI have been used in the ALSPAC, and at age 7 years the opportunity was taken to investigate the specificity

(tendency to identify a nonobese child as obese) and sensitivity (failure to identify an obese child as obese) of these 2 methods against a measure of fatness derived

from arm-to-leg bioimpedance.35 BMI obesity cut-off at the 95th percentile based on the UK 1990 growth refer-

ence106 gave a false-positive rate of 6% (specificity) and a false-negative rate of 12% (sensitivity). The optimum

cut-off with the highest specificity and sensitivity (8% each false-negative and false-positive rates) was the 92nd

percentile. There was no difference between boys and girls in sensitivity and specificity of these cut-offs.

However, when the IOTF cut-offs for obesity (equiva- lent to BMI of 30kg/m2 in adults)110 were used, there

was low sensitivity and a difference in sensitivity be- tween boys and girls, with a false-negative rate of 54%

in boys and 28% in girls (P < 0.01). Specificity was very high and not different between the sexes. Therefore, use

of the IOTF cut-off would likely underestimate obesity prevalence, and this underestimation would be greater

in boys than girls. When IOTF cut-offs for overweight (equivalent to BMI of 25–30 kg/m2 in adults) were used,

sensitivity was much higher (3%–10% false-negative rate), but specificity was reduced, particularly in girls.35

It is very important to be clear about the reference data used when comparing data from different studies,

and these data suggest that in this cohort, at least, the use of the UK 1990 growth reference may be more

informative. A separate analysis investigated whether measuring

waist circumference might provide a better method of assessing fatness than measuring BMI.36 Data collected

using dual x-ray absorptiometry (DXA) at age 9 years was used to assess fat mass, and high fat mass was deter-

mined as being in the top decile of fat mass in each sex separately. Receiver operator characteristics were as- sessed for BMI and waist circumference z-scores. The

area under the curve and specificity were slightly higher for BMI than waist circumference in both sexes (all

P > 0.05).36 Thus, waist measurements have no advan- tage over BMI for the detection of fatness, at least at age

9 years. From bioimpedance measurements (leg-to-leg)

collected at 4 ages between age 7 years and age 11 years an index for lean and fat mass was calculated using re-

siduals from linear regression models in >7000 chil- dren at each age.37 These indices incorporated the

height and age of the child at measurement and were calculated separately for each sex. To test whether these

indices had any functional meaning, their relationship

with cardio-respiratory fitness and grip strength was as-

sessed. In both sexes at age 9 years, lean mass index but not FMI (z-score) was associated with cardio-

respiratory fitness (partial correlation coefficient [r] for lean mass index: in boys, 0.20 [95% CI, 0.15–0.25]; in

girls, 0.26: [95% CI, 0.22–0.30]). BMI z-score showed a slightly weaker relationship with cardio-respiratory fit- ness (in boys, 0.11 [95% CI, 0.06–0.16]; in girls, 0.17

[95% CI, 0.13–0.21]). Very similar relationships were found with grip strength at age 11 years. These results

suggest that bioimpedance-derived lean mass index has functional significance and that FMI, thus derived, is

more likely to be a measure of nutritional status than of function.

The reliability of DXA and bioimpedance measures of fat mass were tested in a study of 176 children aged

11 years in the ALSPAC whose body fatness was mea- sured using DXA and bioimpedance. These results were

compared with those from labeled deuterium dilution, a gold-standard method.38 Fat mass from deuterium dilu-

tion measurements differed from fat mass calculated for the other methods: in boys, fat mass from deuterium di-

lution was 9.8 (SD, 6.1) kg; for DXA, the bias was þ0.9 (limits of agreement, �2.2 to þ 4.1 kg); for bioimpe- dance, the bias was �5.2 kg (limits of agreement, �10.8 to þ0.5 kg): in girls, fat mass from deuterium dilution was 12.1 (SD, 7.7) kg; for DXA, the bias was þ1.2 (lim- its of agreement, �1.9 to þ 5.1 kg); for bioimpedance, the bias was �0.2 kg (limits of agreement, �5.5 to þ5.1 kg). There were also differences in slope and inter- cept between the methods. The limits of agreement and regression analysis suggested that errors in assessment

of fat mass can be very large by either DXA or bioimpe- dance when compared with deuterium dilution and can

differ between the sexes.38 However, the study con- cludes that both of these methods provide adequate esti-

mates of relative fatness for use in large studies such as the ALSPAC.

Childhood growth. The timing of growth measured by

weight gain was investigated to ascertain when excess weight gain was most likely to occur.39 Standardized

measurements were available at various ages between birth and 15 years in 625 children. At birth, mean

weight z-score was 0.12 (SD, 0.97) in comparison with the UK 1990 growth reference data. The mean weight z-

score had increased to 0.22 (SD, 1.03) at age 12 months and remained near to this level at each age through

7 years. By age 9 years, it had increased to 0.37 (SD, 1.03) and by age 11 years to 0.54 (SD, 1.03), staying

around this level at each age through 15 years. The mean BMI z-score was very similar to the weight

z-score from age 12 months (0.20 [SD, 0.99]) through age 5 years but showed a reduction at age 7 years to

190 Nutrition ReviewsVR Vol. 73(S3):175–206

0.13 (SD, 1.06), followed by an increase at age 9 years to

0.34 (SD, 1.10), which remained through age 15 years.39

The weight gain data suggest there is a second period of

fast weight gain beyond infancy that occurs after age 7 years and before age 11 years. The BMI data demon-

strate the occurrence of an adiposity nadir prior to age 7 years and a rebound in the majority of children after age 7 years and before age 9 years.

A separate analysis with the main aim of compar- ing weight gain in the 5% of infants who grew slowly

in the first few months of life with the rest of the co- hort (95%) used all available data at each age and cal-

culated conditional weight and length/height gain.40

This confirmed a period of fast growth in infancy in

the normally growing part of the cohort (condi- tional gain z-score from age 2 months to age

9 months for weight, 0.17 [95% CI, 0.14–0.19], n¼10992; for length, 0.20 [95% CI, 0.17–0.24], n¼8638), then maintenance of weight status up to age 7 years and fast growth between age 7 years and age

10 years (for weight, 0.53 [95% CI, 0.49–0.56], n¼5301; for height, 0.44 [95% CI, 0.41–0.48] n¼5283), slowing slightly between age 10 years and age 13 years (for weight, 0.34 [95% CI, 0.30–0.39],

n¼4539; for height, 0.38 [95% CI, 0.34–0.42], n¼4527). These 2 analyses confirm the later growth spurt after age 7 years and before age 11 years, and the BMI trajectory analysis42,51, corroborates that, on av-

erage, BMI starts to increase (the adiposity rebound) at approximately age 7 years. This may be a time when

interventions aimed at reducing excessive weight gain in children would be effective.

Birth weight and ponderal index at birth were posi- tively associated with lean body mass at age 9 years in

both sexes in 7336 ALSPAC children with DXA mea- surements of body composition.41 Both were also posi-

tively associated with fat mass, such that the association with birth weight was equivalent to a 2%–3% increase

in fat mass per 1 SD increase in birth weight and the as- sociation with ponderal index was an approximately 7% increase in fat mass per 1 SD increase in ponderal in-

dex. These associations were adjusted for gestational age, current height, and SEB. The results suggest that

the well-known relationship between birth weight and BMI reflects increases in both lean and fat tissue and

that ponderal index at birth is a better predictor of adi- posity than birth weight. A further investigation used

ponderal index in the first 2 years and BMI from age 2 years to age 10 years in a longitudinal analysis using

random-effects linear spline models and having DXA-assessed fat mass at 15 years as the outcome

(n¼4601).42 Increases over time of both ponderal in- dex and BMI were associated with greater fat mass at

age 15 years (Table 7). The period of BMI change most

strongly associated with fat mass at age 15 years was ages 2–5 years for girls and ages 5–5.5 years for boys

(Table 7). Associations with cardiovascular risk factors were strongest for BMI change between age 8.5 years

and age 10 years and were largely mediated by fat mass at age 15 years and not evident for ponderal index

changes from age 0 to age 2 years. In girls only, at age 9 years an analysis investigated

associations between infancy weight gain and later growth.43 In linear regression models, current weight,

Table 7 Ponderal index/body mass index trajectories from birth to age 10 years and their association with Ln of DXA-assessed total body fat mass at age 15 years, with multiple imputations, adjusted for age, previous periods of ponderal index/body mass index change, and confounders Ponderal index/ BMI change period

Logged DXA-assessed fat mass (95% CI) a

Boys (n¼2181) Ponderal index at birth 0.038 (�0.004–0.079) Ponderal index

change 0–2 m 0.109 (0.067–0.152)

Ponderal index change 2–24 m

0.164 (0.048–0.281)

BMI change 2–5 y 0.172 (0.134–0.209) BMI change 5–5.5 y 0.558 (0.511–0.605) BMI change 5.5–6.5 y 20.259 (20.298 to 20.220) BMI change 6.5–7 y 20.491 (20.612 to 20.370) BMI change 7–8.5 y 0.446 (0.343–0.549) BMI change 8.5–10 y 0.232 (0.165–0.299) Girls (n¼2420) Ponderal index at birth 0.093 (0.054–0.131) Ponderal index

change 0–1 m 0.102 (0.058–0.146)

Ponderal index change 1–4 m

0.218 (0.171–0.264)

Ponderal index change 4–24 m

0.200 (0.126–0.273)

BMI change 2–5 y 0.306 (0.271–0.342) BMI change 5–5.5 y 0.301 (0.264–0.337) BMI change 5.5–6.5 y �0.244 (20.286 to �0.203) BMI change 6.5–7 y �0.101 (�0.178 to �0.024) BMI change 7–8.5 y 0.222 (0.180–0.263) BMI change 8.5–10 y �0.063 (�0.139 to �0.012) Reproduced from Howe LD, et al. (2010)42 with permission. a All variables were standardized prior to analysis. Coefficients represent the SD change in DXA-assessed total body fat mass associated with a 1 SD increase in the rate of ponderal index/ BMI change. Bold text indicates that ponderal index/BMI is, on average, declining during that period. Adjusted for age, previous periods of ponderal index/BMI change, height, height squared, sex, maternal and partner education, house- hold social class, maternal age, height, gestational age at birth, maternal and partner BMI, maternal and partner smok- ing during pregnancy, age at clinic attendance, and pubertal stage at measurement of outcome. Ponderal index/BMI change periods: BMI change 2–5 years: 24 and 60 months for boys, 24 and 56 months for girls. BMI change 5–5.5 years: 60 and 65 months for boys, 56 and 67 months for girls. BMI change 5.5–6.5 years: 65 and 75 months for boys, 67 and 73 months for girls. BMI change 6.5–7 years: 75 and 81 months for boys, 73 and 79 months for girls. BMI change 7–8.5 years: 81 and 103 months for boys, 79 and 105 months for girls. BMI change 8.5–10 years: 103 and 120 months for boys, 105 and 120 months for girls. Abbreviations: BMI, body mass index; CI, confidence interval; DXA, dual-energy X-ray absorptiometry.

Nutrition ReviewsVR Vol. 73(S3):175–206 191

height, and BMI were positively associated with weight

gain in infancy between age 0 and age 2 months, age 2 months and age 9 months, and age 9 months and age

19 months (all P < 0.01). However, FMI was only asso- ciated with weight gain in the first 2 periods (regression

coefficients adjusted for height at 9 y and maternal edu- cation, 0–2 mo: b¼0.15 6 0.06 [P¼0.01]; 2–9 mo, b¼0.09 6 0.04 [P¼0.001]). In logistic regression anal- ysis, each 1 U increase in weight SDS between age 0 and age 9 months was associated with a 48% increased risk

of overweight at age 9 years (OR, 1.48; 95% CI, 1.27– 1.60), again confirming the importance of rapid growth

in the first year of life.43

Secular trends in growth were investigated using

waist circumference measurements taken in the CIF 4 times between age 2.5 years and age 5.0 years (from

1995 to 1998) compared with measurements taken in a cross-sectional sample of children of similar ages in

1987, up to 11 years earlier.44 BMI was also available for all of the children. Four age groups were compared

with the sexes kept separate; boys waists were 0.60– 1.99 cm larger in the mid-1990s than in 1987 (a 1.2%–

4.1% difference), and girls waists were 1.34–2.50 cm larger (a 2.6%–5.2% difference; all P < 0.05). In both

sets of children, waist circumference increased with age; however, in both sexes in 1987 and the mid-1990s,

BMI declined between age 2 years and age 5 years— e.g., in boys in the ALSPAC, at age 2.5 years, the mean

BMI was 16.7 (SD, 1.3) kg/m2 and at 5.0 years, it was 16.0 (SD, 1.4) kg/m2. This is consistent with the adipos-

ity rebound occurring after age 5.0 years in the majority of children. At each age (except at age 5 y in boys), BMI

was slightly higher in the mid-1990s than in 1987 (in boys, 1.4%–1.8% higher; in girls, 1.6%–4.1% higher; all

P < 0.05).44 These data suggest that in the period be- tween 1987 and 1995–1998 central fatness in children,

measured by waist circumference, had increased and this increase was slightly faster than the concurrent in-

crease in BMI.

Obesity development throughout childhood. The first published report that looked at obesity prevalence in

the ALSPAC was from the CIF subsample at age 5 years.45 Higher frequency of obesity (7.2%) and over-

weight (18.7%) were found than expected when com- pared with the 95th and 85th percentiles of BMI,

respectively, using the UK 1990 growth reference data (both P > 0.001). The incidence of obesity over the time

span 3–15 years was investigated using the CIF subsam- ple from age 3 years to age 15 years (n¼549) and the whole cohort from age 7 years to age 15 years (n¼4283) with the same definition of obesity as above.46 The 4-year incidence of obesity was highest be- tween age 7 years and age 11 years (6.7% in CIF; 5.0%

in the whole cohort), with incidence slightly lower be-

tween age 3 years and age 7 years (5.1% in CIF; data for whole cohort not available) and very much lower be-

tween age 11 years and age 15 years (1.6% in CIF; 1.4% in the whole cohort). In the CIF substudy, the risk of

the child being overweight or obese at age 15 years was 2.4 (95% CI, 1.8–3.1) times higher if the child was over- weight or obese compared with a healthy weight at age

3 years, 4.6 (95% CI, 3.6–5.8) times higher for the same at age 7 years, and 9.3 (95% CI, 6.5–13.2) times higher

for the same at age 11 years.46 There were similar results in the whole cohort. These results suggest a cumulative

effect on obesity development over childhood. A separate analysis (with the sexes separated) inves-

tigated whether a child who was overweight (BMI z-score, �1.04 but <1.64 using the UK 1990 growth ref- erence) at age 7 years was likely to progress to being obese (BMI z-score, �1.64) at age 13 years (n¼5175).47 Children who were overweight at age 7 years were much more likely (38% and 30% in boys and girls, re-

spectively) than normal-weight children (both 5%) to become obese by age 13 years (adjusted OR for boys,

20.5 [95% CI, 12.6–33.6]; for girls, 16.4 [95% CI, 10.0– 27.0]). More than 68% of children who were obese at

age 7 years remained so at age 13 years.47 Adjustment for SEB and parental obesity did not attenuate these re-

lationships to a meaningful extent. Fat mass indices from bioimpedance were used to

define excess fatness (using internal standards) at age 7 years and age 11 years, and excess fatness was tracked

over time (n¼6066) in comparison with BMI-derived overweight and obesity (defined using IOTF cut-

offs).48,110 Table 8 shows a very high degree of tracking in the normal-weight and fatness categories, such that

children who were of normal fatness or BMI at age 7 years were very unlikely to become very over-fat or

obese by age 11 years. The higher fatness categories were less stable than the higher BMI categories

(Table 8) and very similar in boys compared with girls.48 At age 7 years, 11% of children had a parent who was obese (prepregnancy), and these children were

more likely to be overweight (21%) or obese (10%) at age 7 years compared with children of normal-weight

parents (15% and 3%, respectively). Over time, BMI tracked more strongly than fat and lean indices and

seemed to reflect tracking of fat and lean mass equally. Taken together, these studies suggest that adoles-

cent obesity has its roots in early and mid-childhood and is well-established before the age of 11 years.

Risk factors for obesity. The risk factors for obesity were investigated in the whole cohort at age 7 years by multi-

variable analyses using binary logistic models.49 There was no significant difference in frequency of obesity

192 Nutrition ReviewsVR Vol. 73(S3):175–206

between boys (9.2%) and girls (8.1%). In the fully ad-

justed models for risk of obesity (n¼5893 with com- plete data), which included maternal education, birth

weight was positively linearly associated (OR, 1.05: 95% CI, 1.03–1.07; P < 0.001), as was smoking during preg-

nancy (OR for 1–9 cigarettes/d, 1.76; 95% CI, 1.21–2.52; P < 0.01), both parents obese prepregnancy (OR, 10.44;

95% CI, 5.11–21.32; P < 0.001), child watching televi- sion for >8 hours per week at age 3 years (OR, 1.55;

95% CI, 1.13–2.12; P < 0.01), and nighttime sleep of <10.5 hours at age 2.5 years (OR, 1.45; 95% CI, 1.10–

1.89; P < 0.01). Breastfeeding and age at commence- ment of complementary feeding, although associated in

the minimally adjusted model, were not robust to full adjustment. In the CIF substudy (n¼909), in fully ad- justed analysis, rapid early growth (highest quartile of

weight SDS at 8 months compared with the rest [OR, 3.13; 95% CI, 1.43–6.68; P¼0.004) and early adiposity rebound (very early rebound compared with later; OR, 15.00; 95% CI, 5.32–42.30; P < 0.001) were indepen-

dently associated with the risk of obesity.49 These data confirm that early life is a critical time in the develop-

ment of mid-childhood obesity and that maternal and childhood lifestyle factors also play a part. These results

reiterate that parental obesity is a very important deter- minant of childhood obesity.

It is likely that genetic variation is driving some

of the association of parental obesity with offspring obesity. In this regard, the investigation of the FTO

gene has been a priority, and the ALSPAC data were used in a multicohort investigation of age-dependent

associations between variations in the FTO gene and BMI from early infancy to age 13 years.50 BMI was modeled using the LMS method,107 and median curves

showed that carriers of the minor alleles had a lower BMI in infancy, an earlier adiposity rebound, and a

higher BMI in later childhood. The allele effects were additive in that BMI curves of children with one minor

allele were in-between those with 2 and those with none. Longitudinal analysis confirmed the cross-

sectional analysis. These results provide important new insights in the role of the FTO gene in obesity

development.

Inequalities in growth and obesity development. There

are socioeconomic disparities in the United Kingdom for childhood growth and obesity development, and

the age at which these start to emerge has been investi- gated using ALSPAC data. Height trajectories from

age 0 to age 10 years were modeled (n¼12 366) across 4 maternal education categories, with the high educa- tional attainment category split into A-level and de-

gree groups.51 There was a clear positive gradient in birth length by maternal education category, and the

differences in height had widened only very slightly by 10 years of age; the mean difference between the low-

est and highest maternal education category at 10 years was 1.4 cm for boys and 1.7 cm for girls and

was in proportion to the birth length differences. Trajectories of ponderal index at ages 0–2 years and

BMI at ages 2–10 years were also investigated using maternal educational attainment categories.52 There

was little evidence of patterning of ponderal index, but differences in BMI trajectory between maternal educa-

tion groups started to emerge at 4 years and widened with increasing age. By age 8 years, there was a clear

gradient in girls; the higher the educational attainment of the mother, the lower the BMI trajectory. For boys

at age 8 years, only those with mothers within the highest educational attainment differed from the other

3 groups. At age 10 years, the mean BMI difference be- tween the highest and lowest maternal education cate-

gories was 0.89 kg/m2 for girls and 0.38 kg/m2 for boys.52

Inequalities in fat mass assessed by DXA at age 9 years and in blood pressure measured at age 10 years

were also assessed using maternal education catego- ries.53 There were differences between boys and girls in

fat mass associations confirmed by the interaction term (P¼0.0052). The slope index, highest to lowest

Table 8 Persistence of obesity, overweight, and normal weight (defined by International Obesity Task Force cut-offs) and fatness categories (measured by bioimpedence) in children from age 7 years until age 11 years IOTF categories All

(n¼6066) Boys (n¼3008)

Girls (n¼3058)

Of total obese at age 7 y (n¼180), percentage at age 11 y Still obese 75 82 70 Overweight 23 16 28 Normal weight 1.7 1.4 1.9

Of total overweight at age 7 y (n¼762), percentage at age 11 y Obese 16 23 11 Still overweight 63 60 66 Normal weight 21 17 24

Of total normal weight at age 7 y (n¼5214), percentage at age 11 y Obese 0.5 0.5 0.4 Overweight 11 11 11 Normal weight 89 89 88

Fatness categories Total 5933 2938 2995

Of total very over-fata at age 7 y (n¼270), percentage at age 11 y Still very over-fat 57 61 54 Over-fat 30 30 30 Normal fat 13 9.1 16

Of total over-fatb at age 7 y (n¼582), percentage at age 11 y Very over-fat 12 16 9 Over-fat 33 32 34 Normal fat 55 52 57

Of total normal fat at age 7 y (n¼5081), percentage at age 11 y Very over-fat 1.0 1.2 0.7 Over-fat 5.6 5.5 5.6 Normal fat 94 93 94

Reproduced from Wright C, et al. (2010)48 with permission. a >95th internal percentile for fat z-score.

b 85th–95th internal percentile for fat z-score.

Nutrition ReviewsVR Vol. 73(S3):175–206 193

maternal educational attainment for the geometric

mean of fat mass, adjusted for height (null¼1.0) was 1.21 (95% CI, 1.08–1.36) for boys and 1.34 (95% CI,

1.23–1.46) for girls. As found for BMI trajectories, the difference in fat mass for girls showed a gradient over

all of the maternal education groups, whereas for boys only those with mothers with the highest educational at- tainment were different. There was some evidence of

inequality in blood pressure in these children (higher systolic and diastolic blood pressure in children with

mothers in the lowest compared with the highest educa- tional attainment); however, this was partly mediated

by adiposity.53 Taken together, these findings suggest that interventions to prevent obesity are particularly

necessary in the children of less-educated mothers and should start early in childhood.

Diet in relation to growth and obesity development

Timing of adiposity rebound. The association of diet

with the timing of adiposity rebound, when BMI starts to rise after a nadir in early childhood, was investi-

gated in the CIF substudy using BMI measurements up to age 5 years (n¼772 with dietary information).54 Children with very early rebound, at or before age 3.5 years (n¼53 [6.9%]), were compared with those with an early rebound between ages 4 and 5 years (n¼156 [20.0%]) and those with a rebound after age 5 years

(n¼563). There was no difference in BMI z-scores be- tween the groups at age 3.5 years, but at age 4 years

and age 5 years, those with very early rebound had much higher mean BMI z-score than those in the other groups (P < 0.001). For children with very early adi-

posity rebound, a higher proportion (23.3%) had at least 1 obese parent (prepregnancy) compared with

children with early (16.4%) or later adiposity rebound (10.4%; P < 0.01 and P < 0.05, respectively). Timing of

adiposity rebound was not independently related to maternal education. There were no dietary differences

between the groups at either age 8 months or age 1.5 years, even after adjustment for body weight.54

However, although food records were collected at age 3.5 years and age 5 years, these have not yet been used

to compare between adiposity rebound groups, and be- cause the early rebound occurs after age 3.5 years and

before age 5 years, it is likely that investigating diet at these ages would be much more informative.

Energy intake during childhood. The association be-

tween parental BMI and offspring BMI may have a die- tary component as well as the known genetic

components. An analysis of diet trajectory using data from the FFQ and diet diaries between age 3 years and

age 13 years and its association with maternal

prepregnancy BMI as a determinant and with child

BMI at age 15 years as the outcome (n¼4197) was per- formed.19 Greater maternal prepregnancy BMI was as-

sociated with greater child EI at age 3 years; mean predicted EI increased by 4 kcal (95% CI, 3–5) per unit

increase in maternal BMI (P < 0.001) after adjustment for maternal age, education, parity, and social class. EI between age 3 years and age 7 years was increased by 1

(95% CI, 0–1) kcal/year per unit increase in maternal BMI (P < 0.001). There was a slower rate of increase be-

tween age 7 years and age 13 years with an increment of �1 (95% CI, �1 to 0) kcal/year per unit increase in ma- ternal BMI (P¼0.04). Predicted EI at age 3 years and change in EI during childhood was associated with

child’s BMI at age 15 years.19 For every 10-kcal increase in EI at age 3 years, there was a 0.19 (95% CI, 0.17–

0.20) kg/m2 increase in BMI at age 15 years (P < 0.001). After additional adjustment for previous EI, for every

10-kcal/year increase in EI between age 3 years and age 7 years, there was an increase in BMI at age 15 years

(0.46; 95% CI, 0.35–0.57 kg/m2; P < 0.001). However, there was no association between change in EI at ages

7–13 years and BMI at age 15 years in the fully adjusted model (�0.13; 95% CI, �0.91 to 0.67 kg/m2; P¼0.75). The total adjusted association between maternal pre- pregnancy BMI and child BMI at age 15 years was 0.33

(95% CI, 0.30–0.36) kg/m2 per unit increase in maternal BMI (P < 0.001), and 18% (95% CI, 16.7%–19.4%) of

this was accounted for by EI from age 3 years to age 13 years (P < 0.002).19

These results imply that the strong association shown in the ALSPAC between maternal BMI and

offspring obesity is partially mediated by early and mid- childhood diet. For example, this suggests that an aver-

age 3-year-old child would consume approximately 40 kcal/day more if their mother had a BMI of 30 kg/m2

than if she had a BMI of 20 kg/m2; this is equivalent to 2–3 sweets/chocolates or 1 sweet biscuit a day.

Dietary energy density and obesity. It is possible that eat- ing a highly energy-dense diet may predispose to in-

creasing fatness over time. This was investigated in the CIF subsample by characterizing dietary energy density

(without drinks) at age 5 years and age 7 years and looking at fat mass determined by DXA at age 9 years

as the outcome.26 This timeframe covers, in part, the period (ages 7–11 y) when the incidence of obesity is at

its highest in the ALSPAC.46 Children with excess adi- posity at age 9 years were defined as those in the top

20% of the log FMI distribution, and their dietary in- takes were compared with those of the rest of the chil-

dren. There was no evidence of an association between dietary energy density at age 5 years and excess adipos-

ity at age 9 years in minimally or fully adjusted models

194 Nutrition ReviewsVR Vol. 73(S3):175–206

(OR, 1.12; 95% CI, 0.90–1.40; n¼459). However, there was a positive association between energy density at age

7 years and excess adiposity at age 9 years in a mini- mally adjusted model (OR, 1.18; 95% CI, 1.04–1.34),

and this was strengthened in the fully adjusted model, particularly by adjusting for misreporting status (OR,

1.36; 95% CI, 1.09–1.69; n¼584). Other confounders included in the fully adjusted models were sex, total EI,

energy from drinks, fat intake, fiber intake, maternal BMI (prepregnancy) and education, overweight status of the child at baseline (by BMI), and TV watching at

age 4.5 years (hours per day by questionnaire).26

Dietary energy density was available again at age 10

years, and body fatness was assessed by calculating FMI at age 13 years from DXA measurements; in this analy-

sis, the relationship between fatness and FTO genotype was also investigated.27 There was no evidence of a rela-

tionship between energy density in the diet at age 10 years and fat mass at age 13 years when controlling for

height and sex only, but after adjustment for misreport- ing of EI, an association emerged (0.21kg; 95% CI,

0.12–0.30 increase in FMI for each kJ/g increase in die- tary energy density; n¼5527). There was no evidence of an interaction between dietary energy density and

the FTO gene in relation to FMI. Regression analysis with both energy density and FTO in the model con-

firmed their independent relationship with change in FMI (Figure 1) (n¼4318). The addition of weight sta- tus at age 10 years to the model attenuated both effect sizes but did not abolish the relationships (Figure 1).

For energy density, this suggests that overweight chil- dren are more likely to consume energy-dense diets,

and this supports a causal role for dietary energy den- sity in increasing fat mass over time. For FTO, the at- tenuation suggests that the effect of FTO genotype on

fatness is cumulative over time.27

Dietary intakes were investigated directly in rela-

tion to FTO genotype at age 10 years in 3589 children with plausible EI to assess whether the FTO locus might

affect appetite.55 There was an association of the minor allele with increased dietary energy and fat intake even

after adjustment for current BMI: total fat consumption increased by 1.5 g/day (P¼0.02 for each allele), and total energy consumption increased by 25 kJ/day (P¼0.03 for each allele). This provides some evidence of a direct effect of the FTO locus on food intake, sug- gesting that its effect on fatness may be, at least partially,

mediated by diet.

Figure 1 Predicting fat mass at age 13 years from dietary energy density (DED) at 10 years and FTO genotype in children (n¼2275). Reproduced from Johnson et al. (2009)27 with permission. Values are regression coefficients and 95% CIs. A statistically significant effect is indicated by a 95% CI that does not include 0. Separate basic models contain either FTO or DED and are adjusted for height at age 13 years and sex; misreporting of EI is included for models with DED. Joint basic model contains FTO and DED and is adjusted for height at age 13 years, sex, and misreporting of energy intake. Adjusted model includes FTO and DED adjusted for height at age 13 years, sex, puberty (Tanner stages 1–5) at age 13 years, misreporting of EI (under-, plausible-, or over-reporter), EI from drinks (kJ/d) at age 10 years, maternal ed- ucation (none, vocational, O level, A level, or degree), TV watching at age 8 years (<1/1–2/2 h/d), physical activity at age 11 years (counts per minute). Adjusted model including overweight status at age 10 years contains the same variables as adjusted model in addition to over- weight status at age 10 years (defined by IOTF criteria). Abbreviatons: CI, confidence interval; EI, energy intake; IOTF, International Obesity Task Force.

Nutrition ReviewsVR Vol. 73(S3):175–206 195

Fast foods tend to be energy dense and a cross-

sectional investigation was carried out into whether ad- olescents who frequently ate in fast-food outlets had a

higher BMI than those who did not.56 Structural equa- tion models were used with data collected from the par-

ent and child FFQs at age 13 years and weights and heights collected in the research clinic at age 13 years (n¼3620). An analysis of types of foods eaten showed that increased frequency of eating at fast-food outlets was associated with higher consumption of unhealthy

foods at home (chips, burgers, pizzas, pies) (b¼0.29; P < 0.001) and lower consumption of healthy foods

(vegetables and raw fruits) (b¼�1.02; P < 0.001).56 An analysis of frequency of visits to fast-food outlets

showed an association with higher BMI z-score (b¼0.08; P < 0.001). In general, eating frequently in fast-food outlets was a marker of eating a more energy- dense diet.

Drinks in relation to obesity. The relationship between

the drinks consumed at age 5 years and age 7 years and FMI at age 9 years and increase in BMI from age 5 years

to age 9 years was investigated using linear regression in the CIF substudy.29 The analysis focused on sugar-

sweetened beverages, but these drinks accounted for only 15% of drinks consumed and only 3% of total EI;

therefore, the investigation was widened to include milk, diet soft drinks, water, and fruit juice intakes. The

prevalence of overweight, defined by IOTF BMI cut- offs, increased by 5% between age 5 years and age 9

years. There was no evidence of an association between sugar-sweetened beverage intakes at age 5 years

(n¼521) or age 7 years (n¼682) and FMI at age 9 years. Intakes of diet soft drinks at both age 5 years and

age 7 years were positively associated with FMI at age 9 years in unadjusted analysis; however, this association

was removed by adjustment for current BMI at age 5 years and age 7 years. At both ages, diet soft drink con-

sumption was correlated with current BMI,29 so it is possible that parents had provided this type of drink to their children in the hope of limiting weight gain. The

intakes of sugar-sweetened beverages in the ALSPAC were very similar to those in the NDNS 1997,92 where

they contributed 3% and 4% of energy in children aged 4–6 years and children aged 7–10 years, respectively,

and in the NDNS, as in the ALSPAC, consumption of diet soft drinks was greater than consumption of sugar-

sweetened beverages. The relationship between milk intake at age 10 years

and later weight and fat mass gain was investigated in a subset of the children who had recorded their diet for a

full 3 days at both age 10 years and age 13 years and had data on physical activity measured by accelerometer

at age 13 years (n¼2270).57 Two separate analyses were

performed on drinks, 1 on total milk intake and the

other on flavored milk intake. All dairy foods were then combined for a further analysis.

For total milk, boys consumed more than girls (254 g/d compared with 193 g/d, respectively; P < 0.001)

at age 10 years.57 For both sexes, approximately two- thirds of this was reduced-fat milk. There was some evi- dence that milk intake was negatively associated with

body fat (%) measured at age 11 years in the fully ad- justed model (P¼0.03), but there was no association with body fat (%) measured at age 13 years. There was, therefore, no evidence from this study that milk intake

plays a role in relation to adiposity. Flavored milk was consumed by 380 (16.7%) of the

children at age 10 years, and there was no difference be- tween consumers and nonconsumers in the percentage

of children who were overweight/obese at age 10 years or in maternal educational attainment or maternal pre-

pregnancy BMI.58 Flavored milk consumers had a higher average intake of energy, fat, carbohydrate, pro-

tein, and calcium but lower intake of fiber compared with nonconsumers. This amounted to almost 150 kcal

extra EI per day for flavored milk consumers, but there was no difference in the amounts of other foods eaten,

including fruits and vegetables. Change in weight and body fat was measured between age 11 years and age 13

years. In children of normal weight at age 10 years, the change in weight and body fat was similar between fla-

vored milk consumers and nonconsumers. However, in overweight/obese children there was some evidence that

the expected decline in body fat (%) was less in con- sumers (mean decline, �1.27%; 95% CI, �2.32 to �0.22) than in nonconsumers (mean decline, �2.90%; 95% CI, �3.42 to �2.37; P¼0.007). This difference was attenuated but not removed by adjustment for pubertal status, maternal characteristics, and other foods in the

diet.58 These results provide some limited evidence that overweight/obese children may benefit from avoiding

flavored milk drinks. When all dairy foods were combined, there was no

association with excess fat accumulation, and the trend

was in the direction of a negative association, particu- larly with full-fat dairy food intake.59 The data do not

suggest that the consumption of dairy foods, in the amounts consumed in the United Kingdom, is related

to obesity development.

Diet and growth in relation to biochemical outcomes

Iron status. Iron status was investigated in relation to food and nutrient intakes at 1.5 years in the CIF sub-

sample (n¼796).60 Ferritin concentration (a measure of stored iron) was more sensitive to nutrient intakes

(energy-adjusted) than hemoglobin; calcium intake was

196 Nutrition ReviewsVR Vol. 73(S3):175–206

negatively related to ferritin (r¼�0.22; P < 0.001), and iron, vitamin C, and fiber intakes were positively related (all P < 0.05). The only robust relationship of nutrients

with hemoglobin was positive with vitamin C intake (r¼0.13; P¼0.001). For particular foods, there was a negative association between the amount of cows’ milk consumed and ferritin (r¼�0.25; P < 0.001) and a pos- itive association between fruit (P¼0.024) and vegetable consumption (P¼0.030) and hemoglobin.60 Associations between fat intake (% energy) quartiles

and hemoglobin and ferritin concentrations were also investigated (n¼666).22 There were no associations with hemoglobin, but ferritin concentration fell as fat intake quartile increased; 4.7% of children in the lowest

quartile had a very low blood ferritin concentration compared with 14.8% in the highest (chi-squared

P¼0.002). Both iron and vitamin C intakes were lower and calcium intakes higher in the highest compared

with the lowest fat intake quartiles (P < 0.001), and these differences may account for the lower ferritin con-

centration found.60 Because the strongest association between nutrient intake and ferritin at age 1.5 years was

a negative one with calcium intake (P < 0.001), it is pos- sible that this is due to the presence of calcium in the

gut adversely affecting the absorption of iron from the diet.114 Cows’ milk is the main source of calcium in the

diet at this age (61% of calcium), and the amount of cows’ milk consumed was negatively associated with

ferritin (P < 0.001). In the top quartile of fat intake, half of the 1.5-year-olds were consuming >500 g/day (20 oz/

d) of cows’ milk. These data suggest that limiting cows’ milk intake to a maximum of 500 g/day (20 oz/d) in

toddlers is sensible.

Blood lipids. The relationships of birth weight, current size, and central obesity with blood cholesterol and tri-

glyceride concentrations at age 2.5 years (n¼385) and age 3.5 years (n¼470) in white singletons were investi- gated.61 Lipids were measured in nonfasting blood sam- ples; therefore, although associations with cholesterol are likely to be only marginally compromised, associa-

tions with triglyceride should be treated with caution. Total cholesterol and triglyceride fell and high-density

lipoprotein (HDL) cholesterol rose between age 2.5 years and age 3.5 years (all P < 0.001); sex differences in

HDL cholesterol were found at age 3.5 years, with boys having higher concentrations. Height was negatively as-

sociated with triglyceride and total cholesterol concen- trations at age 2.5 years in both sexes (P ranging from

0.091 to 0.016), but at age 3.5 years the only association evident was negative with triglycerides in girls

(P¼0.010).61 The only association between BMI and lipid concentrations was negative with low-density lipo-

protein (LDL) cholesterol in both sexes at age 3.5 years.

Regression associations with birth weight were con-

trolled for current height and BMI, central obesity, and breastfeeding history, and the only independent associ-

ation was negative with HDL cholesterol in boys at age 3.5 years (b¼�0.093 [standard error, 0.03]; P¼0.007). Central obesity was assessed by the ratio of waist-to- arm circumference, and there were some strong inde- pendent associations with lipid concentrations, using

regression models adjusted for current height and BMI, birth weight, and breastfeeding history. At both ages in

boys, increasing central obesity was associated with in- creasing triglyceride concentration (both P < 0.01).

There was a negative association of central obesity with HDL cholesterol in boys at both ages (P¼0.028 and 0.002, respectively) with a marginal negative association only at age 3.5 years in girls (P¼0.072).61 LDL choles- terol showed a marginal negative association with cen- tral obesity in boys (P¼0.053) and a stronger quadratic association in girls (P < 0.010) at age 2.5 years but no associations at age 3.5 years.61 These results do not sup-

port a relationship between birth weight and blood lipid profile but do suggest that a central distribution of fat

may be adversely related to blood lipid concentrations. Dietary associations with blood lipid concentra-

tions at age 2.5 years were investigated in relation to in- take at age 1.5 years.62 There were some relationships

between diet and total cholesterol in boys (n¼214); as- sociations with energy-adjusted nutrients were positive

with total fat (correlation coefficient r¼0.209; P¼0.002), saturated fatty acids (r¼0.211; P¼0.002), and total sugar (r¼0.152; P¼0.026) and negative with carbohydrates (r¼�0.172; P¼0.012). In regression analysis that included these dietary variables and cur- rent height, the only independent relationships with

cholesterol were with height (P¼0.044) and saturated fat intake (P¼0.001) in boys. Slightly higher total cho- lesterol was associated with consumption of compared with abstinence from particular foods, e.g., biscuits

(P¼0.011) and chocolates (P¼0.012). There were no relationships between diet and HDL cholesterol in boys, and relationships with LDL cholesterol were similar to

but weaker than those with total cholesterol. In girls (n¼133), there was a positive association between EI and HDL cholesterol (r¼0.204; P¼0.018) but no asso- ciation with total or LDL cholesterol.62 The positive re-

lationship with energy was independent of other nutrients in the regression model (P¼0.020), with some evidence of a negative relationship with polyun- saturated fatty acids (P¼0.036) emerging.

Associations between dietary fat intake (% energy) quartiles at 1.5 and 3.5 years and blood lipid concentra-

tions at 2.5 and 3.5 years were examined.22 There was some evidence that total cholesterol in boys (at 2.5 y)

was higher, on average, in the top 2 compared to the

Nutrition ReviewsVR Vol. 73(S3):175–206 197

bottom 2 quartiles of fat intake at 1.5 years (P¼0.006). The relationship in girls was inconsistent. There were no associations between fat intake quartiles at 3.5 years

and total or HDL cholesterol at 3.5 years in either sex.22

Relationships between diet and blood lipid concentra-

tions in these preschool children were inconsistent over time and differed between the sexes.

Insulin-like growth factor. There are several insulin-like growth factors (IGFs) that play a part in regulating

childhood growth, and these have been associated posi- tively with the risk of several cancers and negatively

with the risk of coronary heart disease in later life.63

Circulating concentrations of these hormones may be

related to dietary intakes and may be a mechanism by which diet affects later health outcomes. Associations of

growth and diet with the IGF system in childhood be- fore the start of puberty were investigated. IGF-I con-

centrations were measured in blood collected at age 5 years and age 7–8 years, and height and components of

height (trunk and leg length) were available in the CIF substudy.63 There were >200 boys and >200 girls with

data at each age; in age-adjusted cross-sectional analysis in both sexes and at both ages, current height was

strongly positively associated with IGF-I (all P < 0.002, after adjustment for maternal education, housing ten-

ure, current BMI, and birth weight), leg length was not more strongly associated than trunk length, and birth

weight was negatively associated (P < 0.006). Growth in stature between age 5 years and age 7–8 years and be-

tween age 7–8 years and age 9–10 years in relation to IGF-I was also examined. There were strong associa-

tions between IGF-I and subsequent growth in height in both sexes at both ages (all P < 0.004), with slight at-

tenuation on controlling for confounders. In boys, leg and trunk length were equally associated; in girls, only

trunk length was associated.63 These results do not sup- port the hypothesis that associations between leg length

and adult cancer risk are mediated through childhood IGF-I; it could be that age at onset of puberty (which is earlier in girls) confounds this relationship.

Dietary associations with IGF-I and IGFBP-3 (the main binding protein for IGF-I) were investi-

gated in 521 white singletons at age 7–8 years in the ALSPAC. Intakes of milk and dairy products had been

associated with IGF concentrations in adults previously, and because children are high consumers of these foods,

they were likely candidates to examine.64 There was evi- dence of a sex difference in the relationship between

milk or dairy product intakes and IGF concentrations; in boys only, both IGF-I and IGFBP-3 were higher in

the top compared with the bottom quartile of intake of milk or dairy products. Adjusted analysis suggested that

animal protein intake rather than milk or dairy per se

might be the important factor.64 A separate analysis in-

vestigated the direct relationship between milk or dairy intake and height (z-scores) in these children (n¼744); again, there were statistically supported sex differ- ences (e.g., interaction for sex � dairy products for leg length P¼0.015), with associations occurring only in boys.63 The association with dairy products was stronger with leg than trunk length. In multivariate

analysis, adjustment for IGF-I attenuated the associa- tion to the null, suggesting that IGF-I mediates this

association. A further analysis examined other aspects of diet in

relation to IGF concentrations.65 An age- and sex-ad- justed analysis confirmed that protein, and in particular

animal protein, was strongly positively associated with IGF-I (P < 0.001) and also with the ratio of IGF-I to

IGFBP-3, which is a crude indicator of IGF-I bioavail- ability (P¼0.003). Positive associations for zinc, phos- phorus, magnesium, calcium, and potassium with IGF-I were also found (all P < 0.008). However, in a regres-

sion model including protein and the 5 minerals listed above, only the association with protein was robust to

adjustment (P¼0.020).65 There were negative associa- tions between intakes of total fat, monounsaturated

fatty acid, and polyunsaturated fatty acid and IGF-I (all P < 0.03), which were stronger in girls than in boys.

IGFBP-3 was positively associated with EI (P¼0.002); other associations were much weaker and differed by

sex. There were no associations with intakes of particu- lar food groups, including meat, processed meat, and

vegetables. Excluding under-reporters from the analysis did not change these associations; however, a positive

association between EI and IGF-I emerged (fully ad- justed P¼0.030).65 These 2 studies found important positive associations between animal protein intake and IGF-I; this was particularly important for dairy pro-

tein, which provides 24% of the total protein intake in these children. These results are consistent with obser-

vations in adults. There was also some evidence of a negative relationship between fat intake and IGF-I, which is consistent with the association of high fat and

low IGF-I with the risk of coronary heart disease in adults.

Insulin. Low birth weight, rapid weight gain in the early

years, shorter adult stature, and lower IGF-I concentra- tions have been shown to be associated with the risk of

developing type 2 diabetes in adulthood.115 The oppor- tunity was taken to investigate whether similar associa-

tions with growth parameters were present in ALSPAC children at age 7–8 years (n¼851) by using fasting in- sulin concentration and insulin secretion (30 min after an oral glucose load had been ingested) as a proxy for

diabetes risk.66 Insulin and glucose concentrations

198 Nutrition ReviewsVR Vol. 73(S3):175–206

(fasting and at 30 min) were used to calculate insulin

sensitivity (a measure of how sensitive the body is to the effects of insulin) and insulin secretion (the amount of

insulin produced in response to a glucose load). Girls had lower insulin sensitivity than boys (P < 0.0005),

and this was not explained by body size. Insulin sensi- tivity decreased with increasing current weight, waist circumference, and BMI (all P < 0.0005). Height was in-

versely related to insulin secretion (P < 0.0005). Early rapid weight gain was associated with lower insulin sen-

sitivity and elevated BMI at 8 years.66 Current BMI ac- counted for 10.2% of the variation in insulin sensitivity,

and early weight gain accounted for 2.1%. There was no association between insulin sensitivity and birth weight

except in children in the highest tertile for current BMI, whereby an inverse association was found (P¼0.0006). Insulin secretion correlated negatively with insulin sen- sitivity and was higher with elevated BMI and with

rapid early weight gain (all P < 0.001).66 There were in- dependent positive relationships of ponderal index at

birth (adjusted P¼0.01) and childhood height (ad- justed P¼0.047) with insulin secretion; thus, thinness at birth and smaller stature were related to reduced in- sulin secretion. IGF-I at 5 years (n ¼252) predicted height gain between age 5 years and age 8 years (P¼0.008) and was positively related to insulin secre- tion at age 8 years (P < 0.001); this association was inde- pendent of current BMI (P¼0.004). The results are in line with findings in adult studies. The mechanisms for these relationships may be related to b-cell function and need further investigation, as does the contribution of diet.

Free fatty acids. The possibility that high concentrations of free fatty acids in the blood of the ALSPAC children

may be associated with poor b-cell function, which could have manifested as reduced insulin secretion,67 was in-

vestigated. This was done in parallel with an adult study in which similar data were available. Higher fasting free

fatty acid concentrations were associated with lower in- sulin secretion in both sexes in both groups (boys:

P¼0.03; girls: P¼0.001; men: P¼0.03; women: P¼0.04). In the adults studied, higher fasting free fatty acid concentrations were associated with greater risk of

developing type 2 diabetes in the following 5–8 years. Again, dietary intakes may contribute to these relation-

ships, and the ALSPAC has the potential to investigate this in the future.

Diet and growth in relation to other childhood outcomes

Birth weight associations with childhood diet. Low birth

weight is associated with cardiovascular disease in later

life. It is possible that birth weight might be related to

differences in childhood diet, and these differences might mediate the association.68 The diet in the CIF

substudy was investigated at age 8 months, 1.5 years, 3.5 years, and 7 years, and there was only minimal evi-

dence of an association between a child’s birth weight and their later dietary intake; a small inverse association with saturated fat intake only at age 3.5 years was abol-

ished in adjusted models. This suggests it is unlikely that diet plays a mediating role in the relationship be-

tween birth weight and later disease, at least in this cohort.

Puberty. Early onset of menarche in girls was associated

with fast early growth; menarche age <12 years was as- sociated with faster weight gain from age 0 to age 2

months (P¼0.006) and age 2 months to age 9 months (P < 0.001) but not with weight gain after that.43

Associations with length gain started after age 2 months and persisted to age 19 months. In logistic regression

analysis, each 1 SD increase in weight SDS between age 0 and age 9 months was associated with a 34% increased

risk for menarche at age <12.0 years (OR, 1.34; 95% CI, 1.21–1.49).

The relationship between dietary intake at age 3 years and age 7 years (by FFQ) and age 10 years (by

food record) and age at menarche was investigated.69

The association found with EI at age 10 years was not

robust to adjustment for current body size; however, re- lationships with protein intake at both age 3 years and

age 7 years did persist (OR for reaching menarche by 12 y 8 mo per 1 SD increase in energy-adjusted protein

intake at 3 y, 1.11; 95% CI, 1.00–1.23; P¼0.057 and OR at 7 years, 1.14; 95% CI, 1.04–1.26; P¼0.007). Meat in- take appeared to be driving the association with protein at both age 3 years and age 7 years. For example, an

early menarche was more likely in girls with an intake of >12 portions of meat compared with <4 portions

of meat per week at age 7 years (OR, 1.57; 95% CI, 1.03–2.37; P¼0.020).69 The analyses were adjusted for maternal education and smoking, maternal age at men-

arche, parity, duration of breast feeding, and birth weight. These results suggest that high protein intake in

early and mid-childhood may affect the timing of puberty.

The relationship between body composition at age 11 years and pubertal stage at age 12 years was investi-

gated in boys and girls using BMI z-scores (UK 1990 reference standards) and bioimpedance-derived lean

and fat indices (internal standardization).48 For girls, early pubertal development (breast development and

pubic hair) was associated with higher mean BMI, fat, and lean mass index z-scores (BMI, 0.70 [95% CI,

0.59–0.79]; FMI, 0.23 [95% CI, 0.12–0.33]; lean mass

Nutrition ReviewsVR Vol. 73(S3):175–206 199

index, 0.35 [95% CI, 0.26–0.45]). For boys, the associa-

tion of early pubertal development (genitalia and pubic hair) with BMI z-score was variable, and lean mass in-

dex z-score showed a tendency to be higher (P < 0.05), whereas FMI z-score showed a tendency to be lower

(P < 0.01).48 Thus, it is important to study boys and girls separately, particularly at the time of puberty.

Tasting ability. Bitter tasting ability was assessed in 4178 10-year-old participants who also had genetic data

available relating to the TAS2R38 locus, which had pre- viously been shown to be associated with bitter tasting

ability.70 A taste score was measured and found to be lower in the group predicted to be nontasters by their

genetic haplotype (median score, 3.7; interquartile range, 5.2) compared with the group predicted to be

tasters (median score, 8.1; interquartile range, 2.5). A further analysis investigated how bitter tasting ability

might relate to childhood growth and feeding behav- iors.71 It is possible that children who are tasters might

be choosier about what they eat than nontasters. Mothers had been asked about whether their child was

choosy with food at several ages from 15 months to 4.5 years. More than half of the children were considered

choosy by their mothers at 15 months, and the propor- tion increased with age. There was no evidence of a dif-

ference between tasting groups even at age 4.5 years when 78% of nontasters compared with 82% of “super”

tasters were said to be choosy (P¼0.10). It may be that if a more strict definition of “being choosy” had been

used, a significant difference would have been found. There were no differences in BMI at age 10 years be-

tween tasting groups but slightly more of the super tast- ers than the nontasters were below the 10th percentile

for height at age 10 years (12% compared with 8%, re- spectively; P¼0.008).71 These data warrant further exploration.

Autism spectrum disorders. The feeding behaviors, diet, and growth of children with autism spectrum disorders (ASDs) were investigated using the FFQs up to age 4

years.72 There were 79 children with ASDs, and they were compared with 12 901 control subjects. Children

with ASDs started solid foods slightly later than control subjects and were more likely to be described as slow

feeders at age 6 months. From age 15 months to age 4 years, they were consistently more likely to be described

as difficult to feed (pooled OR, 2.92; 95% CI, 2.08–4.09; P < 0.001) and very choosy about food (pooled OR, 2.55;

95% CI, 1.91–3.40; P < 0.001). Nutrient intakes estimated from the FFQ at age 3 years did not differ between chil-

dren with ASDs and control subjects, except those with ASDs had slightly lower intakes of vitamin C and ate less

variety of foods, particularly consuming fruit and

vegetables less often. There were no differences between

children with ASDs and control subjects in weight, height, or BMI at age 18 months or age 7 years.72 These

findings on eating behaviors are consistent with the idea that children with ASDs have difficulty accepting

change.

DISCUSSION

This review illustrates the great advantage that the

ALSPAC has over cross-sectional studies of dietary in- take and childhood growth because it follows the same

children through childhood to adolescence with regular dietary assessments and measurements of size. The

ALSPAC is able to identify when important changes in diet and growth occur. These valuable insights can in-

form decisions about the targeting of interventions to improve diet and encourage optimal growth, thus gain-

ing the most efficient use of scarce resources. The con- tribution of longitudinal birth cohorts to scientific

knowledge is now widely recognized.116 The ALSPAC has been in the vanguard of this type of research and

has shown that including some measurements not im- mediately driven by a particular hypotheses can lead to

important findings, e.g., those relating to iron status or autism. For each section of the results, the value added

by assessing the publications from the ALSPAC in one review are discussed with the key findings, in brief,

listed in Box 1. The longitudinal comparison of the nutrient and

food group intakes over childhood in the ALSPAC clearly show that the greatest change in dietary intake oc-

curs during the preschool period15,16 and that the diet re- mains relatively stable after that (Table 4).15–18 There is a

step change in free sugars intake at this time from ap- proximately 12% to approximately 16% of the energy in-

take. This is within the context of the recommendation of a maximum intake of 10% of energy from free sug-

ars93 and is also seen cross-sectionally in the NDNS 2008–2012.102 Contributions of protein to EI decline slightly at the same time, and total EI rises in line with

the size of the child (Table 4).15–18 Throughout child- hood, average fiber intake is at only 75% of the amount

thought to be adequate (Table 4). The increase in intake of sugars is likely due to increased numbers of con-

sumers of energy-dense, noncore foods such as sugar confectionery, sweetened breakfast cereals, puddings,

and ice cream (Table 5).15–18,23 The proportion of con- sumers of these types of foods remain similar at each

succeeding age of assessment. The contribution to energy intake of these foods rises by 11% during this preschool

period (Table 6).23 These results suggest that preschool could be a key time to provide a healthy eating environ-

ment for children and to educate parents about diet.

200 Nutrition ReviewsVR Vol. 73(S3):175–206

Overall, the diets of children in the ALSPAC were adequate for most nutrients, with the notable exception

of vitamin D. At age 1.5 years and age 3.5 years, all of the ALSPAC children were found to be below the

United Kingdom dietary recommendations for vitamin D intake, and children who had low intakes at age

1.5 years were twice as likely to have low intakes at age 3.5 years.24 Although it has been recommended for

25 years or more that children aged <5 years should be given vitamin D supplements,93 only a minority in ei- ther the ALSPAC24 or the national surveys actually

were supplemented.101 It is possible that a fortification program could be a good option to improve this situa-

tion. It was also evident that relative intakes of other important nutrients such as calcium and vitamin A de-

clined in the preschool years,23 and this was due to the children starting to consume more energy-dense, nutri-

ent-poor foods (Table 6). Fruit and vegetable intakes were below recommen-

dations at all ages in the ALSPAC children, and low in- takes were established before starting school (Table 5).

The in-depth investigation of determinants of fruit and vegetable intake at age 7 years 28 confirmed that boys ate

less than girls and that both family and child characteris- tics were important in determining intake. The ALSPAC

results add to the evidence that mothers act as role mod- els and gatekeepers for their children regarding healthy

eating by showing that children whose mothers ate more fruit and vegetables and had rules about serving fruit and

vegetables daily had greater intakes.28 Other studies have also found that the diet of the mother is a key influence

on the diet of the child.117,118 In the ALSPAC, the child’s own eating behaviors, particularly choosiness with food,

affected intake, especially of vegetables. Further support is needed to educate mothers about healthy food choices for their children and to encourage parents to eat well

themselves. The need to offer advice and help to parents was

manifest in the investigation of school meal quality.33

The packed lunches brought from home by the children

were almost always inadequate compared with recom- mendations,93,99 and other meals eaten during the day

did not fully compensate for this.33 The cooked meals provided by the school were slightly better nutritionally

than the packed lunches. Food-based standards for pri- mary school meals were introduced in the United

Kingdom in 2006, and a systematic review of studies as- sessing meals taken by children in schools throughout

the United Kingdom has found an improvement in the provided school dinners but no improvement in packed

Box 1 Key findings in brief

• Diets in childhood changed decisively in the preschool years toward greater energy density. Increased intake of

free sugars was the main reason for the greater energy density of the diet.

• Low intakes of vitamin D were common throughout childhood. At preschool ages, all ALSPAC children had die-

tary intakes below the UK dietary recommendations. Uptake of vitamin D supplements was low, in line with

national surveys and despite recommendations that all children aged 1–5 years should receive them.

• Fruit and vegetable intakes were low at all ages; both family and child characteristics were important in deter-

mining intake.

• Maternal educational attainment was related to the quality of the child’s diet throughout childhood; lower edu-

cation was associated with a more energy-dense diet with less fruit and vegetables consumed.

• There were 2 periods during which faster-than-average weight gain predisposed to obesity: early infancy and

between age 7 years and age 11 years, when most children experience the adiposity rebound.

• Having 1 or both parents who were obese prior to pregnancy was one of the strongest risk factors for being

obese in childhood. Genetic and dietary factors had associations that were independent of each other.

• Children who were overweight or obese in early or mid-childhood were much more likely than normal-weight

children to be obese as adolescents.

• There were associations between the energy density of the diet in mid-childhood and the development of obe-

sity at the time of the adiposity rebound.

• Levels of fatness in girls increased stepwise as maternal educational attainment decreased, but in boys, only

those whose mothers had the highest educational attainment were slimmer than the rest.

• Interventions to prevent obesity are particularly necessary in the children of less-educated mothers; they should

start at preschool age and include dietary change.

• Interventions to improve diet quality should aim to decrease intakes of sugary, energy-dense foods and in-

crease intakes of fruit and vegetables. Parental involvement is key to success in improving children’s diets.

Nutrition ReviewsVR Vol. 73(S3):175–206 201

lunches.119 Uptake of school dinners has increased since

the time of the ALSPAC assessment, but quality stan- dards must be maintained to maximize the benefit.

Misreporting of food intake, particularly by adoles- cents, is very likely in dietary assessment and has been

found in many studies.120 The ALSPAC dietary studies have confirmed that as children progress to adoles- cence, the level of likely under-reporting of EI increases

(Table 4)15,16,24,26 and that overweight or obese individ- uals are more likely to under-report their intake than

normal-weight individuals.18,31 It is important to con- sider when assessing relationships between diet and var-

ious outcomes that misreporting affects all aspects of the diet due to the high correlation of energy with all

macronutrients and most micronutrients. Furthermore, the ALSPAC data showed there was a bias toward the

under-reporting of particular types of foods, often non- core and energy-dense foods. It is very difficult to find

dietary assessment methods that are applicable in a pop- ulation survey that are not subject to misreporting, and

the ALSPAC results endorse the necessity of using sta- tistical methods to account for misreporting to avoid

misleading conclusions.26,27

The various analyses carried out using ALSPAC

data suggest that in large-scale studies BMI provides a useful method (relatively cheap and easy) of assessing

body size35,38; however, its use for measuring obesity may be more problematic. The use of waist circumfer-

ence to assess obesity did not appear to have any advan- tage over BMI.38 Using BMI to determine obesity and

overweight identified different individuals depending on the growth reference data applied.34,42 For example,

when IOTF cut-offs for obesity were used, false-positive rates differed by sex, but there was no difference by sex

when using the 95th percentile of the UK 1990 growth reference standards.42 Several analyses of ALSPAC data

confirm the benefit to be gained from assessing fat and lean body mass when studying growth and obesity.36,37

As with BMI, the different methods used were shown to identify different individuals as having excess fat.36

Although leg-to-leg bioimpedance is a cheap and easy

method, there are difficulties in finding a suitable equa- tion to calculate fat mass from bioimpedance.37 Using a

DXA scan may give a slightly more precise measure- ment of fat and lean mass,36 but the equipment is ex-

pensive and not very portable, making this method impractical for most field work. These reports from the

ALSPAC emphasize the importance of considering the method and growth reference data used to assess obe-

sity when comparing results across studies. Several analyses using ALSPAC data have examined

growth trajectories and confirm 2 periods of fast growth: early infancy and between age 7 years and age 11

years.39,42 The latter is the time of increasing BMI known

as the adiposity rebound. Investigations relating weight

and ponderal index at birth to growth found positive as- sociations with BMI and both fat and lean mass; how-

ever, ponderal index at birth was a better predictor of adiposity in adolescence than birth weight.41 Following

from this, increases in ponderal index between age 0 and age 2 years and BMI between age 2 years and age 10 years were associated with greater fat mass at age 15 years

(Table 7).51 ALSPAC data has also substantiated a secular trend toward increasing BMI over time.44 Furthermore,

these studies highlight differences in growth rate between boys and girls (Table 7).44,51

The development of obesity throughout childhood has been the theme of several investigations of ALSPAC

data, which identified that the highest incidence of obe- sity (defined by BMI) was between age 7 years and age

11 years; it was slightly lower between age 3 years and age 7 years and very low during adolescence.46 BMI is a

measure of both fat and lean mass and tends to track through childhood. ALSPAC data showed that fat mass

varies more over time than BMI (Table 8). Children who were of normal BMI in mid-childhood were very

unlikely to become obese by the start of adolescence, whereas those who were already overweight or obese

were likely to remain so.47,48 These results suggest that interventions to reduce obesity in schools would only

be cost effective if targeted at overweight or obese chil- dren. There were very similar findings and conclusions

in a large, nationally representative study of children in the United States who were followed from age 5 years

to age 14 years between 1998 and 2007.121

When risk factors for mid-childhood obesity were

investigated, rapid early growth and early adiposity re- bound were found to be important predictors.49 Early

adiposity rebound occurred in 7% of children at around age 3.5 years and in 20% between 4 and 5 years54 and

was much more likely if either parent was obese (pre- pregnancy). Having one or both parents who were

obese (prepregnancy) was a strong predictor for mid- childhood obesity.49 The genetic underpinning of this association was shown in investigation of the FTO ge-

notype, where having the minor allele was associated with increasing BMI between age 7 years and age

11 years.50 On the diet side, greater maternal prepreg- nancy BMI was associated with greater child EI at age

3 years and rate of increase of EI up to age 7 years.19

Furthermore, the association of mothers’ BMI with off-

spring BMI continued into adolescence and was par- tially explained by the child’s EI in early childhood.19

The detailed dietary data collected in the ALSPAC during mid-childhood facilitated investigation of the as-

sociation between the energy density of the diet and the development of obesity at the time of its highest inci-

dence in these children.46 As children progressed into

202 Nutrition ReviewsVR Vol. 73(S3):175–206

adolescence, diets with high energy density were in-

creasingly associated with higher dietary EI.28,63 Greater energy density at age 7 years was associated with an in-

creased risk of adiposity at age 9 years,28 with similar findings for energy density at age 10 years and fat mass

at age 13 years (Figure 1).63 An independent association with fat mass at age 13 years was found for the FTO ge- notype of the child (Figure 1).63 These studies confirm

that both diet and genetic make-up are important in obesity development.

Differences in diet associated with SEB of the family, particularly in relation to maternal educational attain-

ment, were found; children of mothers with low educa- tion consume less fiber (nonstarch polysaccharide) and

more free sugars with more noncore and fewer core foods, particularly fruit, than children of mothers with

high education.18,20 It is probable that these dietary differ- ences have a role to play in the difference in adiposity be-

tween children of mothers with different extents of educational attainment. Variations in childhood adiposity

between education groups began to emerge by age 4 years.51–53 By late childhood, fatness in girls increased

stepwise as maternal educational attainment decreased, but in boys, only those with mothers with the highest ed-

ucational attainment were slimmer than the rest.52

Interventions to prevent obesity are particularly necessary

in the children of the less-educated mothers; they should start early in childhood and include dietary change.

ALSPAC data were used to investigate relationships between diet and growth and biochemical markers of

health. In the children at preschool ages, the relation- ship between diet and iron status was explored.60

Ferritin was negatively associated with calcium and cows’ milk intake and positively associated with iron in-

take, and hemoglobin was positively associated with vi- tamin C and fruit and vegetable intake. At each age of

dietary assessment, many children, particularly girls, had low dietary iron intakes. Dietary advice should aim

to maximize iron intake and absorption. Further analyses examined associations between

food and nutrient intakes and the IGF system and found,

in particular, a positive relationship between IGF-I and animal protein intake.64,65 There were strong positive as-

sociations between IGF-I and growth in height.63 Diet- induced variations in the IGF axis in childhood may

have implications for the long-term risk of several chronic diseases. Exploration of insulin profiles in rela-

tion to growth showed that high BMI in mid-childhood and rapid weight gain in infancy were associated with a

less favorable insulin profile in mid-childhood.66 These findings were in line with adult studies and show the im-

portance of monitoring childhood growth and exploring the relationship of faster-than-average growth in infancy

and childhood with health indicators.

Strengths and limitations

One of the major strengths of this study is the availabil-

ity of detailed records of foods/drinks as consumed at

many ages throughout childhood in a cohort of pro-

spectively followed children. The ability to explore de-

tails of diet that this data provides is extremely valuable

and cannot easily be matched by other studies. Further

strengths are the careful collection of anthropometric

data by trained staff, mostly annually, and the inclusion

of more precise ways of measuring body composition as

they became available and affordable—these include

bioimpedance measures and DXA scan measures.

These methods are able to assess fat and lean mass and

have shown that fat mass is a more sensitive measure of

obesity development in relation to dietary intake than

BMI. The availability of prospectively collected con-

founding variables, as well as details of the pregnancy

and infancy of the child, adds further value to this co-

hort. However, the timing of data collection has not al-

ways been ideal due to delays in obtaining ethical

approval and/or funding. It has also been important to

keep participant burden to a minimum.

There are several limitations that apply to all obser-

vational studies, and one is that causality cannot be proved. In some analyses, new statistical methods, such

as Mendelian randomization, which uses knowledge of the genetic makeup of the cohort, has been used to help

overcome this problem. Another inevitable conse- quence is incomplete data collection and follow-up

when cohort members do not take part in all phases of the study or decide to drop out completely. The ALSPAC has made particular efforts over the years to

retain the cohort and has been relatively successful73; however, only approximately half of the full cohort pro-

vided food records at age 7 years, and there was further attrition at later ages (Table 2). For some analyses, sta-

tistical methods have been used to account for missing data, and most have aimed to take account of the biases

that may distort the results by adjusting for SEB. The ALSPAC is a geographically defined cohort, albeit rea-

sonably representative of the population at recruitment; the results, therefore, need to be evaluated from this

perspective. Members of the cohort were born in 1991– 1992, and there have been many changes in life experi-

ence for children in the United Kingdom since that time; therefore, the results always need to be seen in

this context. For growth and obesity development, there are reference data that can be used to set ALSPAC data

in context. This is also true of dietary data because in the United Kingdom there is a rolling program of die-

tary assessment in nationally representative cross- sectional samples covering all age groups at least every

10 years. Thus, ALSPAC dietary data can be compared

Nutrition ReviewsVR Vol. 73(S3):175–206 203

with dietary data collected at similar ages in similar and

different eras so that changes in diet over time can be evaluated. The food and nutrient intakes of children at

each age in ALSPAC have been compared with this data and shown to be very similar, suggesting that ALSPAC

results may be generalizable in the United Kimgdom.

CONCLUSION

The longitudinal assessment of diet in ALSPAC’s child-

hood cohort has identified a critical time of dietary change between the ages of 1.5 years and 3 years when

diet moves decisively toward increased intake of free sugars and, thus, increased dietary energy density.

Dietary energy density showed a strong association with increasing body fatness between the ages of 7 years

and 9 years and between the ages of 9 years and 13 years once adjustments were made for misreporting of EI.

These analyses suggest that the most promising area for dietary intervention in relation to obesity prevention is

the manipulation of dietary energy density. The inter- vention should start during preschool and aim to sup- port parents in continuing good dietary habits

throughout childhood. The dietary changes should aim to reduce the energy density of the diet and include

foods that have a high nutrient density. To this end, the foods to encourage are core foods such as fruits and

vegetables of all types, plain potatoes, pasta and rice, high-fiber/low-sugar breakfast cereals, high-fiber bread,

plain meat, and fish. Consumption of noncore foods, including sweets and chocolate confectionery,

sweetened breakfast cereals, white bread, sweet biscuits, and fried potatoes, should be discouraged. It would be

beneficial if the whole family were encouraged to eat in this way because mothers are influential as both role

models and providers of food to children. It would be prudent to design food provision in preschools and

schools to reinforce these healthier food habits and to work with the food industry to promote the recom-

mended foods.

Acknowledgments

The authors are extremely grateful to all the families

who took part in this study, the midwives for their help in recruiting them, and the whole ALSPAC team, which

includes interviewers, computer and laboratory techni- cians, clerical workers, research scientists, volunteers,

managers, receptionists, and nurses. Author contributions. This publication is the work of

the authors, and P.M.E. serves as guarantor for the con- tents of this paper. P.M.E. drafted the article and was

involved in interpretation. L.R.J. contributed to the

drafting of the article and the interpretation of the data. Both authors approved the final version.

Funding. The UK Medical Research Council (Grant ref:

74882) and the Wellcome Trust (Grant ref: 092731) and

the University of Bristol provide core support for

ALSPAC. This review was specifically funded by Wyeth

Nutrition but was carried out independently.

Declaration of interest. The work was supported by

Wyeth Nutrition, although carried out independently.

P.M.E. and L.R.J. have from time to time received re-

search funding, and P.M.E. has received consultancy

funding from Pfizer Nutrition Ltd., and Danone Baby

Nutrition (Nutricia Ltd.). PME currently receives re-

search funding from Nestlé Nutrition.

REFERENCES

1. Joint World Health Organization and Food and Agriculture Organization of the United Nations Expert Consultation. “Diet, nutrition and the prevention of chronic diseases.” World Health Organ Tech Rep Ser 916, no. i-viii 2003.

2. Steyn NP, Mann J, Bennett PH, et al. Diet, nutrition and the prevention of type 2 diabetes. Public Health Nutr. 2004;7:147–165.

3. World Cancer Research Fund, American Institute for Cancer Research. Food, Nutrition and the Prevention of Cancer: a Global Perspective. Washington, DC: American Institute for Cancer Research; 1997.

4. Mikkila V, Rasanen L, Raitakari OT, et al. Consistent dietary patterns identified from childhood to adulthood: the Cardiovascular Risk in Young Finns Study. Br J Nutr. 2005;93:923–931.

5. Berenson GS, Wattigney WA, Bao W, et al. Epidemiology of early primary hyper- tension and implications for prevention: the Bogalusa Heart Study. J Human Hypertens. 1994;8:303–311.

6. Berenson GS, Srinivasan SR, Nicklas TA. Atherosclerosis: a nutritional disease of childhood. Am J Cardiol. 1998;82:22T–29T.

7. Maynard M, Gunnell D, Emmett P, et al. Fruit, vegetables, and antioxidants in childhood and risk of adult cancer: the Boyd Orr cohort. J Epidemiol Community Health. 2003;57:218–225.

8. Lobstein T, Baur L, Uuay R. Obesity in children and young people: a crisis in pub- lic health. Obes Rev. 2004;5:4–85.

9. Jackson-Leach R, Lobstein T. Estimated burden of paediatric obesity and co- morbidities in Europe, part 1: the increase in the prevalence of child obesity in Europe is iteself increasing. Int J Paed Obesity. 2006;1:26–32.

10. Ogden CL, Carroll MD, Flegal KM. High BMI for age among US children and ado- lescents JAMA. 2008;299:2401–2405.

11. Baird J, Fisher D, Lucas P, et al. Being big or growing fast: systematic review of size and growth in infancy and later obesity. BMJ. 2005;331:929–931.

12. Reilly J. Physical acitivity, sedentary behaviour, and energy balance in the pre- school child: early opportunities for obesity prevention. Proc Nutr Soc. 2008;67:317–325.

13. Kipping RR, Jago R, Lawlor DA. Obesity in children. Part 2: prevention and man- agement. BMJ. 2008;337:a1848.

14. Emmett PM, Jones LR. Diet and growth in infancy: relationship to socioeconomic background and to health and development in the Avon Longitudinal Study of Parents and Children. Nutr Rev. 2014;72:483–506.

15. Cowin I, Emmett P. Diet in a group of 18-month-old children in South-West England, and comparison with the results of a national survey. J Hum Nutr Diet. 2000;13:87–100.

16. Emmett PM, Rogers IS, Symes C, et al. Food and nutrient intakes of a population sample of 3-year- old children in the South West of England in 1996. Public Health Nutr. 2002;5:55–64.

17. Glynn L, Emmett P, Rogers I, et al. Food and nutrient intakes of a population sample of 7-year-old children in the south-west of England in 1999/2000—what difference does gender make? J Hum Nutr Diet. 2005;18:7–19.

18. Cribb VL, Jones LR, Rogers IS, et al. Is maternal education level associated with diet in 10-year-old children? Public Health Nutr. 2011;14:2037–2048.

19. Anderson EL, Tilling K, Fraser A, et al. Estimating trajectories of energy intake through childhood and adolescence using linear-spline multilevel models: the Avon Longitudinal Study of Parents and Children. Epidemiology. 2013;24:507–515.

20. Rogers I, Emmett P. The effect of maternal smoking status, educational level and age on food and nutrient intakes in preschool children: results from

204 Nutrition ReviewsVR Vol. 73(S3):175–206

the Avon Longitudinal Study of Parents and Children. Eur J Clin Nutr. 2003;57:854–864.

21. Rogers I, Emmett P,. Fat content of the diet amond preschool children in south- west Britain; relationship with food and nutrient intakes. Eur J Clin Nutr. 2002;56:236–244.

22. Rogers I, Emmett P,. Fat content of the diet among preschool children in south- west Britain. II. Relationship with growth, blood lipids and iron status. Pediatrics. 2001;108:e49.

23. Cribb VL, Northstone K, Hopkins D, et al. Sources of vitamin A in the diets of pre- school children in the Avon Longitudinal Study of Parents and Children (ALSPAC). Nutrients. 2013;2013:1609–1621.

24. Cribb V, Northstone K, Hopkins D, et al. Sources of vitamin D and calcium in the diets of preschool children in the UK and the theoretical effect of food fortifica- tion [published online ahead of print October 3, 2014]. J Hum Nutr Diet. doi/ 10.1111/jhn.12277

25. Sayers A, Fraser W, Lawlor D, et al. 25-Hydroxyvitamin-D3 levels are positively re- lated to subsequent cortical bone development in childhood: findings from a large prospective cohort study. Osteoporosis Int. 2012;23:2117–2128.

26. Johnson L, Mander AP, Jones LR, et al. A prospective analysis of dietary energy density at age 5 and 7 years and fatness at 9 years among UK children. Int J Obesity. 2007;31:1–8.

27. Johnson L, van Jaarsveld CHM, Emmett PM, et al. Dietary energy density affects fat mass in early adolescence and is not modified by FTO variants. PLoS One. 2009;4:e4594.

28. Jones LR, Steer CD, Rogers IS, et al. Influences on child fruit and vegetable intake: sociodemographic, parental and child factors in a longitudinal cohort study. Public Health Nutr. 2010;13:1122–1130.

29. Johnson L, Mander AP, Jones LR, et al. Is sugar-sweetened beverage consump- tion associated with increased fatness in children? Nutrition. 2007;23:557–563.

30. Jago R, Ness AR, Emmett P, et al. Obesogenic diet and physical activity: indepen- dent or associated behaviours in adolescents? Public Health Nutr. 2010;13:673–681.

31. Noel SE, Mattocks C, Emmett P, et al. Use of accelerometer data in prediction equations for capturing implausible dietary intakes in adolescents. Am J Clin Nutr. 2010;92:1436–1445.

32. More J, Emmett P. Evidence-based, practical food portion sizes for preschool chil- dren and how they fit into a well balanced, nutritionally adequate diet. J Hum Nutr Diet. 2015; 28:135–154.

33. Rogers IS, Ness AR, Hebditch K, et al. The quality of food eaten in English primary schools: school dinners versus packed lunches. Eur J Clin Nutr. 2007;61:856–864.

34. Wright C, Lakshman R, Emmett P, et al. Implications of adopting the WHO 2006 Child Growth Standard in the UK: two prospective cohort studies. Arch Dis Child. 2008;93:566–569.

35. Reilly J, Dorosty A, Emmett P. Identification of the obese child: adequacy of the body mass index for clinical practice and epidemiology. Int J Obes Relat Metab Disord. 2000;24:1623–1627.

36. Reilly JJ, Dorosty AR, Ghomizadeh NM, et al. Comparison of waist circumference percentiles versus body mass index percentiles for diagnosis of obesity in a large cohort of children. Int J Ped Obesity. 2010;5:151–156.

37. Sherriff A, Wright CM, Reilly JJ, et al. Age-and sex-standardised lean and fat indi- ces derived from bioelectrical impedance analysis for ages 7–11 years: functional associations with cardio-respiratory fitness and grip strength. Br J Nutr. 2009;101:1753–1760.

38. Reilly JJ, Gerasimidis K, Paparacleous N, et al. Validation of dual-energy x-ray ab- sorptiometry and foot-foot impedance against deuterium dilution measures of fatness in children. Int J Ped Obesity. 2010;5:111–115.

39. Hughes AR, Sherriff A, Lawlor DA, et al. Timing of excess weight gain in the Avon Longitudinal Study of Parents and Children (ALSPAC). Pediatrics. 2011;127:e730–e736.

40. ud Din Z, Emmett P, Steer C, et al. Growth outcomes of weight faltering in in- fancy in ALSPAC. Pediatrics. 2013;131:e843–e849.

41. Rogers IS, Ness AR, Steer CD, et al. Association of size at birth and dual-energy X-ray absorptiometry measures of lean and fat mass at 9 to 10 y of age. Am J Clin Nutr. 2006;84:739–747.

42. Howe LD, Tilling K, Benfield L, et al. Changes in ponderal index and body mass index across childhood and their associations with fat mass and cardiovascular risk factors at age 15. PLoS One. 2010;5:e15186.

43. Ong KK, Emmett P, Northstone K, et al. Infancy weight gain predicts child- hood body fat and age at menarche in girls. J Clin Endocrinol Metab. 2009;94:1527–1532.

44. McCarthy H, Jarrett K, Emmett P, et al. Trends in waist circumferences in young British children: a comparative study. Int J Obesity. 2005;29:157–162.

45. Reilly JJ, Dorosty AR, Emmett PM. Prevalence of overweight and obesity in British children: cohort study. BMJ. 1999;319:1039–1039.

46. Hughes AR, Sherriff A, Lawlor DA, et al. Incidence of obesity during child- hood and adolescence in a large contemporary cohort. Prevent Med. 2011;52:300– 304.

47. Reilly JJ, Bonataki M, Leary SD, et al. Progression from childhood overweight to adolescent obesity in a large contemporary cohort. Int J Paed Obesity. 2011;6:e138–143.

48. Wright C, Emmett P, Ness A, et al. Tracking of obesity and body fatness through mid-childhood. Arch Dis Child. 2010;95:612–617.

49. Reilly JJ, Armstrong J, Dorosty AR, et al. Early life risk factors for obesity in child- hood: cohort study. BMJ. 2005;330:1357–1363.

50. Sovio U, Mook-Kanamori DO, Warrington NM, et al. Association between com- mon variation at the FTO locus and changes in body mass index from infancy to late childhood: the complex nature of genetic association through growth and development. PLoS Genet. 2011;7:e1001307.

51. Howe LD, Tilling K, Galobardes B, et al. Socioeconomic differences in childhood growth trajectories: at what age do height inequalities emerge? J Epidemiol Community Health. 2012;66:143–148.

52. Howe L, Tilling K, Galobardes B, et al. P19 Socioeconomic disparities in trajectories of adiposity across childhood. J Epidemiol Community Health. 2010;64(Suppl 1):A41.

53. Howe L, Galobardes B, Sattar N, et al. Are there socioeconomic inequalities in car- diovascular risk factors in childhood, and are they mediated by adiposity? Findings from a prospective cohort study. Int J Obesity. 2010;34:1149–1159.

54. Dorosty AR, Emmett PM, Cowin S, et al. Factors associated with early adiposity rebound. ALSPAC Study Team. Pediatrics. 2000;105:1115–1118.

55. Timpson NJ, Emmett PM, Frayling TM, et al. The fat mass– and obesity-associ- ated locus and dietary intake in children. Am J Clin Nutr. 2008;88:971–978.

56. Fraser L, Edwards K, Cade J, et al. Fast food, other food choices and body mass in- dex in teenagers in the United Kingdom (ALSPAC): a structural equation model- ling approach. Int J Obesity. 2011;35:1325–1330.

57. Noel SE, Ness AR, Northstone K, et al. Milk intakes are not associated with per- cent body fat in children from ages 10 to 13 years. J Nutr. 2011;141:2035–2041.

58. Noel S, Ness A, Northstone K, et al. Associations between flavored milk consump- tion and changes in weight and body composition over time: differences among normal and overweight children. Eur J Clin Nutr. 2013;67:295–300.

59. Bigornia SJ, LaValley MP, Moore LL, et al. Dairy intakes at age 10 years do not ad- versely affect risk of excess adiposity at 13 years. J Nutr. 2014 144:1081–1090.

60. Cowin I, Emond A, Emmett P, et al. Association between composition of the diet and haemoglobin and ferritin levels in 18-month-old children. Eur J Clin Nutr. 2001;55:278–286.

61. Cowin IS, Emmett PM, Team AS. Cholesterol and triglyceride concentrations, birthweight and central obesity in preschool children. Int J Obesity. 2000;24:330–339.

62. Cowin IS, Emmett PM, Team AS. Associations between dietary intakes and blood cholesterol concentrations at 31 months. Eur J Clin Nutr. 2001;55:39–49.

63. Rogers I, Metcalfe C, Gunnell D, et al. IGF-I and growth in height, leg length and trunk length between ages 5 and 10 years: the ALSPAC cohort. J Clin Endocr Metab. 2006;91:2514–2519.

64. Rogers I, Emmett P, Gunnell D, et al. Milk as a food for growth? The insulin-like growth factors link. Public Health Nutr. 2006;9:359–368.

65. Rogers IS, Gunnell D, Emmett A, et al. Cross-sectional associations of diet and IGF lev- els in 7-8 year old children. Cancer Epidemiol, Biomarkers Prevent. 2005;14:204–212.

66. Ong K, Petry C, Emmett P, et al. Insulin sensitivity and secretion in normal chil- dren related to size at birth, postnatal growth, and plasma insulin-like growth factor-I levels. Diabetologia. 2004;47:1064–1070.

67. Salgin B, Ong KK, Thankamony A, et al. Higher fasting plasma free fatty acid levels are associated with lower insulin secretion in children and adults and a higher incidence of type 2 diabetes. J Clin Endocrinol Metabol. 2012;97: 3302–3309.

68. Shultis W, Leary S, Ness A, et al. Does birth weight predict childhood diet in the Avon Longitudinal Study of Parents and Children? J Epidemiol Community Health. 2005;59:955–960.

69. Rogers IS, Northstone K, Dunger DB, et al. Diet throughout childhood and age at menarche in a contemporary cohort of British girls. Public Health Nutr. 2010;13:2052–2063.

70. Timpson NJ, Heron J, Day IN, et al. Refining associations between TAS2R38 diplo- types and the 6-n-propylthiouracil (PROP) taste test: findings from the Avon Longitudinal Study of Parents and Children. BMC Genetics. 2007;8:51.

71. Golding J, Steer C, Emmett P, et al. Associations between the ability to detect a bitter taste, dietary behavior, and growth: a preliminary report. Ann N Y Acad Sci. 2009;1170:553–557.

72. Emond A, Emmett PM, Steer C, et al. Feeding symptoms, dietary patterns, and growth in young children with autistm spectrum disorders. Pediatrics. 2010;126:e337–e342.

73. Boyd A, Golding J, Macleod J, et al. Cohort profile: the “children of the 90s”—the index offspring of the Avon Longitudinal Study of Parents and Children. Int J Epidemiol. 2013;42:111–127.

74. Fraser A, Macdonald-Wallis C, Tilling K, et al. Cohort profile: the Avon Longitudinal Study of Parents and Children: ALSPAC mothers cohort. Int J Epidemiol. 2013;42:97–110.

75. Avon Longitudinal Study of Parents and Children Data Dictionary. University of Bristol Web site. http://www.bris.ac.uk/alspac/researchers/data-access/data-dic- tionary/. Updated March 21, 2015. Accessed 31 August 2015.

76. Rogers I, Emmett P. Diet during pregnancy in a population of pregnant women in South West England. Eur J Clin Nutr. 1998;52:246–250.

Nutrition ReviewsVR Vol. 73(S3):175–206 205

77. Northstone K, Smith ADAC, Cribb VL, et al. Dietary patterns in UK adolescents ob- tained from a dual-source FFQ and their associations with socio-economic posi- tion, nutrient intake and modes of eating. Public Health Nutr. 2013;17:1476–1485.

78. Avon Longitudinal Study of Parents and Children Questionnaires. University of Bristol Web site. http://www.bristol.ac.uk/alspac/researchers/resources-available/ data-details/questionnaires. Accessed 31 August 2015.

79. Crawley H. Food Portion Sizes. London: Ministry of Agriculture Fisheries and Food and HMSO; 1990.

80. Wrieden WL, Longbottom PJ, Adamson AJ, et al. Estimation of typical food portion sizes for children of different ages in Great Britain. Br J Nutr. 2008;99:1344–1353.

81. Price GM, Paul AA, Key FB, et al. Measurement of diet in a large national survey: comparison of computerised and manual coding of records in household mea- sures. J Hum Nutr Diet. 1995;8:417–428.

82. Holland B, Welch AA, Unwin ID, et al. McCance & Widdowson’s the Composition of Foods. 5th ed. Cambridge: Royal Society of Chemistry; 1991.

83. Holland B, Unwin ID, Buss DH. Cereals and cereal products. Third Supplement to McCance & Widdowson’s the Composition of Foods. 4 ed. Cambridge: Royal Society of Chemistry; 1988.

84. Holland B, Unwin ID, Buss DH. Milk products and eggs. Fourth Supplement to McCance & Widdowson’s the Composition of Foods. Cambridge: Royal Society of Chemistry; 1989.

85. Holland B, Unwin ID, Buss DH. Vegetables, herbs and spices. Fifth Supplement to McCance & Widdowson’s the Composition of Foods. 4th ed. Cambridge: Royal Society of Chemistry; 1991.

86. Holland B, Unwin ID, Buss DH. Fruit and nuts. First Supplement to McCance & Widdowson’s the Composition of Foods. 5th ed. Cambridge: Royal Society of Chemistry; 1992.

87. Holland B, Welch AA, Buss DH. Vegetable dishes. Second Supplement to McCance & Widdowson’s the Composition of Foods. 5th ed. Cambridge: Royal Society of Chemistry; 1992.

88. Holland B, Brown J, Buss DH. Fish and fish products. Third Supplement to McCance & Widdowson’s the Composition of Foods. 5th ed. Cambridge: Royal Society of Chemistry; 1993.

89. Chan W, Brown J, Buss DH. Miscellaneous foods. Fourth Supplement to McCance & Widdowson’s the Composition of Foods. 5th ed. Cambridge: Royal Society of Chemistry; 1994.

90. Chan W, Brown J, Lee SM, et al. Meat, poultry and game. Fifth Supplement to McCance and Widdowson’s the Composition of Foods. 5th ed. Cambridge: Royal Society of Chemistry; 1995.

91. Chan W, Brown J, Church SM, et al. Meat products and dishes. Sixth Supplement to McCance & Widdowson’s the Composition of Foods. 5th ed. Cambridge: Royal Society of Chemistry; 1996.

92. Gregory J, Lowe S. National Diet and Nutrition Survey: Young People aged 4 to 18 Years. Volume 1: Report of the Diet and Nutrition Survey. London: Stationery Office; 2000.

93. Department of Health. Dietary Reference Values for Food Energy and Nutrients for the UK. Report of the Committee on Medical Aspects of Food Policy. London: HMSO; 1991.

94. Australian Guide to Healthy Eating. Australian Government, Department of Health Web site. https://www.eatforhealth.gov.au/guidelines/australian-guide- healthy-eating. Accessed 31 August 2015.

95. Rennie KL, Coward A, Jebb SA. Estimating under-reporting of energy intake in di- etary surveys using an individualised method. Br J Nutr. 2007;97:1169–1176.

96. Torun B. Energy requirements of children and adolescents. Public Health Nutr. 2005;2005:968–993.

97. Black AE, Cole TJ. Within- and between-subject variation in energy expenditure measured by the doubly-labelled water technique: implications for validating re- ported dietary energy intake. Eur J Clin Nutr. 2000;54:386–394.

98. Willett WC, Howe GR, Kushi LH. Adjustment for total energy intake in epidemio- logic studies. Am J Clin Nutr. 1997;65(Suppl):S1220–S1228.

99. Caroline Walker Trust. Eating Well at School: Nutritional and Practical Guidelines. Abbots Langley, Hertfordshire: Caroline Walker Trust; 2005.

100. The Eatwell Plate. Public Health England Web site. https://www.gov.uk/govern- ment/uploads/system/uploads/attachment_data/file/340869/2014-250_-_eat- well_plate_Final_version_2014.pdf. Accessed 31 August 2015.

101. Gregory JR, Collins DL, Davies PSW, et al. The National Diet and Nutrition Survey: Children Aged 1 1/2 to 4 1/2 Years. Volume I. Report of the Diet and Nutrition Survey. London: HMSO; 1995.

102. Bates B, Lennox A, Bates C, et al. National Diet and Nutrition Survey. Results from Years 1-4 (combined) of the Rolling Programme (2008/2009 – 2011/12) London: Public Health England and Food Standards Agency; 2014.

103. Riddoch CJ, Mattocks C, Deere K, et al. Objective measurement of levels and pat- terns of physical activity. Arch Dis Child. 2007;92:963–969.

104. de Vries SI, Bakker I, Hopman-Rock M, et al. Clinimetric review of motion sensors in children and adolescents. J Clin Epidemiol. 2006;59:670–680.

105. Mattocks C, Ness A, Leary S, et al. Use of accelerometers in a large field-based study of children: protocols, design issues and effects on precision. J Phys Activity Health. 2008;5 S98–S111.

106. Cole TJ, Freeman JV, Preece MA. Body mass index reference curves for the UK, 1990. Arch Dis Child. 1995;73:25–29.

107. Cole TJ. The LMS method for constructing normalised growth standards. Eur J Clin Nutr. 1990;44:45–60.

108. Wells JC, Cole TJ. Adjustment of fat-free mass and fat mass for height in children aged 8 years. Int J Obes Relat Metab Disord. 2002;26:947–952.

109. Ambrosini GL, Emmett PM, Northstone K, et al. Identification of a dietary pattern prospectively associated with increased adiposity during childhood and adoles- cence. Int J Obes. 2012;36:1299–1305.

110. Cole TJ, Bellizzi MC, Flegal KM, et al. Establishing a standard definition for child overweight and obesity worldwide: international survey. BMJ. 2000;320:1–6.

111. Monteilh C, Kieszak S, Flanders WD, et al. Timing of maturation and predictors of Tanner stage transitions in boys enrolled in a contemporary British cohort. Paediatr Perinatal Epidemiol. 2011;25:75–87.

112. Bellisle F, Drewnowski A. Intense sweeteners, energy intake and the control of body weight. Eur J Clin Nutr. 2007;61:691–700.

113. World Health Organization Multicentre Growth Reference Study Group. WHO child growth standards based on length/height, weight and age. Acta Paediatrica (Oslo, Norway: 1992). 2006;450:76.

114. Hallberg L, Rossander-Hulten L, Brune M, et al. Calcium and iron absorption: mech- anism of action and nutritional importance. Eur J Clin Nutr. 1992;46:317–327.

115. Forsén T, Eriksson J, Tuomilehto J, et al. The fetal and childhood growth of per- sons who develop type 2 diabetes. Anns Int Med. 2000;133:176–182.

116. Lawlor DA, Anderson AM, Batty GD. Birth cohort studies: past, present and fu- ture. Int J Epidemiol. 2009;38:897–902.

117. Fisk CM, Crozier SR, Inskip HM, et al. Influences on the quality of young children’s diets: the importance of maternal food choices. Br J Nutr. 2011;105:287–296.

118. Vereecken CA, Keukelier E, Maes L. Influence of mother’s educational level on food parenting practices and food habits of young children. Appetite. 2004;43:93–103.

119. Evans CEL, Cleghorn CL, Greenwood DC, et al. A comparison of British school meals and packed lunches from 1990 to 2007: meta-analysis by lunch type. Br J Nutr. 2010;104:474–487.

120. Livingstone MBE, Black AE. Markers of the validity of reported energy intake. J Nutr. 2003;133(Suppl):895S–920S.

121. Cunningham SA, Kramer MR, Venkat Narayan KM. Incidence of childhood obesity in the United States. New Engl J Med. 2014;370:403–411.

206 Nutrition ReviewsVR Vol. 73(S3):175–206