Interventions for the prevention of overweight and obesity in preschool children: a systematic review of randomized controlled trials

Authors


Lorenzo Monasta, Epidemiology and Biostatistics Unit, Institute for Maternal and Child Health IRCCS Burlo Garofolo, Via dell'Istria 65/1, 34137 Trieste, Italy. E-mail: monasta@burlo.trieste.it

Summary

The objective of this study was to analyse interventions for the prevention of overweight and obesity in children under 5 years of age. We carried out a systematic review focusing exclusively on randomized controlled trials (RCTs). Data sources include Medline, Cochrane Library, EMBASE, CINHAL, PsychInfo and Web of Science. Data were extracted from seventeen articles describing seven RCTs identified through electronic search, screening of references in systematic reviews, own files and contact with authors. RCTs were assessed with the Jadad scale. Four trials were carried out in preschool settings, one with an exclusive educational component, two with an exclusive physical activity component and one with both. Two trials were family-based, with education and counselling for parents and children. The remaining trial was carried out in maternity hospitals, with a training intervention on breastfeeding. None of the interventions had an effect in preventing overweight and obesity. The failure to show an effect may be due to the choice of outcomes, the quality of the RCTs, the suboptimal implementation of the interventions, the lack of focus on social and environmental determinants. More rigorous research is needed on interventions and on social and environmental factors that could impact on lifestyle.

Introduction

In Western countries, the prevalence of overweight and obesity has been rapidly increasing for more than a decade (1,2). This has paralleled changes in dietary, physical activity (PA) and sedentary activity behaviours, with an unequal distribution across socioeconomic groups (3,4). These secular increases in overweight and obesity represent a major health problem for adults, and an increasing concern for children and adolescents. Although the evidence for a rising trend among children under 5 years of age is not strong (5–8), this is an age in which remarkable physical and metabolic change takes place and the complex web of determinants starts setting the grounds for the subsequent manifestation of obesity (9). In children and young people, the short-term deleterious effects of obesity and overweight on psychological, behavioural and physiological health have been well documented. Physiological responses include an unfavourable impact on blood pressure, blood lipids and blood glucose (including diabetes) (10,11). Given that these characteristics are markers of coronary heart disease (CHD) risk, some investigators have examined the long-term influence of excessive body weight in children and young people on later clinical CHD events. In the few studies identified in a recent systematic review (12), obesity in child and young adult populations, typically measured by body mass index (BMI), was associated with an increased risk of later CHD in most studies.

The first years of life can therefore be crucial to start preventive interventions that can have an impact on lifestyle and on later overweight and obesity. Thus, there is an urgent need to identify effective and sustainable interventions targeting children under 5 years of age, before overweight and obesity occur in later childhood and adolescence, and in order to prevent the consolidation of obesogenic behaviours. Through a systematic review of randomized controlled trials (RCTs), the objective of this report is to describe, analyse and discuss potentially effective interventions for the prevention of childhood overweight and obesity in children under 5 years of age. None of the identified published systematic reviews focused exclusively on RCTs conducted on children under 5 years of age.

Methods

Scoping electronic search and approaches

In March 2008, we conducted a scoping search of Medline to determine the general quantity and quality of the literature on interventions for the prevention of overweight and obesity in children under 5 years of age. The following simple and limited search strategy was used: [(((Intervention*) AND (("Overweight"[Mesh]) OR ("Obesity"[Mesh])) AND (English[lang] OR French[lang] OR Italian[lang] OR Spanish[lang] OR Portuguese[lang]) AND (infant[Mesh] OR child, preschool[Mesh])))]. This identified 405 citations. This preliminary list of titles and abstracts helped us to develop the definitive search strategy and, together with the preliminary review conducted to study the determinants of overweight and obesity (9), also helped us design the conceptual framework.

Our review only focused on RCTs for the prevention of overweight and obesity in children under 5 years of age. We used four approaches to identify relevant RCTs. First, an electronic search was performed to find RCTs and systematic reviews on the prevention of overweight and obesity in children under 5 years of age. Second, the lists of references of the identified RCTs and systematic reviews were scrutinized to identify additional publications. These first two steps were carried out independently by two researchers (LM and AB). Third, we searched our own files. Finally, we contacted experts in the field in an attempt to gather information on the status of ongoing projects we had identified but for which we could not find evidence of results, or otherwise to request additional materials and publications on implemented RCTs.

Full electronic search

The full electronic search was carried out in August 2008. We searched six databases: Medline (from its inception in January 1966); The Cochrane Library (Issue 3, 2008) – including The Central Register of Controlled Trials (CENTRAL) and DARE; EMBASE (from 1980); CINHAL (Cumulative Index to Nursing & Allied Health; from 1982); PsychInfo (from 1887); and Web of Science (from 1970). For Medline we used the search strategy reported in Table 1. This strategy was adapted where necessary to search the other databases (details available upon request). Citations were catalogued and managed with Reference Manager 11 (13). Data were extracted from hard copies of the full papers and the quality of all included manuscripts was independently evaluated by two researchers (LM and AM) using the Jadad scale, a simple, validated, widely used and most cited 0 to 5 score for RCT quality assessment (Fig. S1) (14). Any disagreement, of which there were few, was resolved by discussion. We included all relevant RCTs regardless of language or publication status. Studies and reviews focusing on treatment of overweight and obesity, and those studies not having anthropometric measures either as primary or secondary outcomes were excluded.

Table 1.  Search strategy in Medline*
  • *

    The strategy was amended where necessary to search the other databases.

 1OBESITY [Mesh]
 2WEIGHT GAIN [Mesh]
 3Obesity OR obese OR obesogenic ti, ab
 4‘Weight gain’ OR overweight OR ‘over weight’ OR ‘weight change’ ti, ab
 5(BMI OR ‘body mass index’) AND (gain OR change OR increase*) ti, ab
 61 OR 2 OR 3 OR 4 OR 5
 7((child OR children (ti, ab)) OR CHILD [Mesh] OR infant* OR INFANT [Mesh] OR pediatr* OR ((pre-school OR preschool (ti, ab)) OR CHILD, PRESCHOOL [Mesh]
 86 AND 7
 9OBESITY/PREVENTION AND CONTROL [Mesh]
10PRIMARY PREVENTION {Mesh}
11((primary prevention ti, ab) OR (secondary prevention ti, ab)
12preventive measure* ti, ab
13preventive care ti, ab
14([obesity] AND [prevent* OR treat*]) ti, ab
15DIET THERAPY [Mesh]
16(diets OR diet OR dieting) ti, ab
17EXERCISE [Mesh]
18EXERCISE THERAPY [Mesh]
19HEALTH PROMOTION [Mesh]
20HEALTH EDUCATION [Mesh]
21((Health*) AND (Promot* OR educat*)) ti, ab
22HEALTH POLICY [Mesh]
23NUTRITION POLICY [Mesh]
249–23/OR
25Randomized controlled trial* ti, ab, pt
26Controlled Clinical trial* ti, ab, pt
27RANDOM ALLOCATION [Mesh]
28DOUBLE BLIND METHOD [Mesh]
29SINGLE BLIND METHOD [Mesh]
30PLACEBOS [Mesh]
31COMPARATIVE STUDY. pt
32PROSPECTIVE STUDIES [Mesh]
33CROSS-OVER STUDIES [Mesh]
34CASE-CONTROL STUDIES [Mesh]
35COHORT STUDIES [Mesh]
36REVIEW [Mesh]
37comparison OR compared (ti,ab)
3825–37/OR
398 AND 24 AND 38

Results

Some 9477 titles were identified through the electronic search of the databases (Fig. 1). From an initial screening of titles and abstracts, we selected 24 articles on potentially relevant RCTs. In parallel, the 9477 titles were screened to identify systematic reviews. From a first selection based on titles and abstracts, we identified 32 systematic reviews of potential interest. A second screening, based on full text, shortened the list to 13 relevant systematic reviews from which we extracted 18 potentially relevant articles. After merging the two lists, one identified through a direct screening of the 9477 references and one identified through systematic reviews, and removing duplicates, we obtained 31 potentially relevant articles that were retrieved and the full text scrutinized. This yielded 14 articles describing seven independent RCTs. Table S1 contains the list of the 17 articles that were excluded after full text review. Reviewing the bibliographies of these 14 articles, and after contacting authors, we identified three additional references regarding two RCTs. This final step led us to a final count of 17 articles describing seven separate RCTs; these are listed in Table 2. The number of articles exceeds the number of trials as, given the complexity and duration of some trials, the outcomes of single trials were often analysed in more than one article. Table S2 shows the 13 published systematic reviews with the articles on RCTs taken into consideration.

Figure 1.

Flow of citations for the identification of randomized controlled trials on interventions for the prevention of overweight and obesity in children under 5 years of age.

Table 2.  Overview of included studies: interventions for the prevention of overweight and obesity in infant and preschool children
Citation
Country
[Jadad score]
Design, unitsPopulation (sample size, age, ethnicity)Interventions, settings, duration, follow-up and number of componentsOutcomesSummary of results
  • *

    I, intervention group; C, control group.

  • Head Start is a programme of the US Department of Health and Human Services providing comprehensive education, health, nutrition and parent involvement services to low-income children and their families.

  • BFHI, Baby Friendly Hospital Initiative; BMI, body mass index; CI, confidence interval; PA, physical activity; OR, odds ratio; RCT, randomized controlled trial; SD, standard deviation.

Dennison (15) (2004)
USA
[1]
Brocodile the Crocodile
Cluster RCT
16 preschools
and/or day-care centres
Sample size: 102 children at baseline, 77 children at end of intervention (I: 43, C: 34)*.
Age 3.5–5.5 years.
Mean age 4.0 years.
Ethnicity: primarily Caucasian children (96%).
Preschool-based interactive sessions by an early childhood teacher and a music teacher to reduce television watching and promote healthy eating and PA.
7 × 20 min interactive education sessions (6 weekly sessions and a final session 1 month later) plus complementary materials and suggested classroom activities for staff and parents.
Duration of intervention: 10.3 weeks approximately.
Follow-up: 12 weeks after intervention.
Number of components: 2.
Triceps skin-fold thickness, BMI.
Behaviour change: parent-reported child television and video viewing.
No statistically significant differences in children's growth parameters between groups.
Overall television watching reduced by mean 3.1 h per week in the intervention group vs. increased by 1.6 h per week in the control group, adjusted difference between means, –4.7 h; 95% CI, –8.4 to –1.0 h; P = 0.02.
Viewing television = 2 h d−1 decreased from 33% to 18% (I) vs. increased from 41% to 47% (C), P = 0.046.
Fitzgibbon (16) (2002)
Stolley (17) (2003)
Fitzgibbon (18) (2005)
Fitzgibbon (19) (2006)
USA
[0]
Hip-Hop to Health Jr.
Cluster RCT
24 preschools
Sample size: 810 children (599 parents) at baseline, 631 children at 2-year follow-up.
1st sample: 409 (I: 197, C: 212) at baseline, 300 at 2-year follow-up (I: 146, C: 154).
2nd sample: 401 (I: 202, C: 199), 331 at 2-year follow-up (I: 171, C: 160).
Aged 2–5 years.
Ethnicity: 50% African–American, 44% Latino, 6% multicultural or other.
24 Head Start preschool programmes (1st part: 12 schools, Sept 1999-Jun 2002; 2nd part: 12 schools, fall 2001 and winter 2003).
Healthy eating and exercise intervention (40 min, three times weekly) in 24 preschools by trained early childhood educators. Each session had two main components (i) a 20-min lesson that introduced a healthy eating or exercise concept with an activity; and (ii) 20 min of ongoing PA. In addition, parents in the intervention group received a weekly newsletter including homework assignment.
For each completed assignment parents received a $5 grocery store coupon.
Duration of intervention: 14 weeks.
Follow-up: 1 year and 2 years after intervention.
Number of components: 3.
BMI value and Z score for change.
Behaviour change: parent reported PA (frequency per week and intensity) and television viewing (hours per day) and a single 24 h recall dietary intake questionnaire by registered dietician.
In the first sample: at 1-year follow-up, mean BMI increases were significantly lower for children in the intervention group when compared with controls (0.06 vs. 0.59 kg m−2; difference adjusted to age quartile, baseline value and preschool –0.53 kg m−2; 95% CI –0.91 to –0.14; P = 0.012); at 2-year follow-up, 0.54 vs. 1.08 kg m−2; adjusted difference –0.54 kg m−2; 95% CI –0.98 to –0.10; P = 0.022).
In the second sample, post-intervention changes in BMI and BMI z score were not significantly different between children in the intervention and control groups. At 1-year follow-up, the mean increase in BMI for children and the change in BMI z score in intervention and control preschools did not differ significantly.
Harvey-Berino (20) (2003)
USA and Canada
[0]
Cluster RCT
43 mother/child pairs
43 mother/child pairs at baseline (I: 23, C: 20) from families that had a child between ages 9 months and 3 years, a mother with BMI>25, and if mother agreed to keep all treatment appointments. 40 pairs at end of intervention (I: 20, C: 20).
Age 9–36 months.
Ethnicity: native-American children.
Subjects recruited through media advertisements, day-care centres, the Special Supplemental Nutrition Program for the Women, Infants and Children, self-referral and informal networking in the community.
Intervention: maternal participation in an obesity prevention (maternal training in child feeding and exercise) plus parenting support. The intervention was delivered one-on-one in homes by an indigenous peer educator. Control: parenting support only.
Mothers were given a $25 coupon for a local mall.
Duration of intervention: 16 weeks.
Follow-up: at end of intervention.
Number of components: 2.
Weight and height (weight-for-height z scores and weight-for-height percentile for children).
Behaviour change: dietary intake (3-d food records), PA (measured by accelerometers), parental feeding style, maternal outcome expectations, self-efficacy, and intention to change diet and exercise behaviours.
No difference between groups in the percentage of children classified as ≥85th or ≥95th weight-for-height percentile at either time-point (baseline and 16 weeks). Children in the intervention group decreased in weight-for-height z scores, while children in the control group increased their weight-for-height z scores (–0.27 ± 1.1 vs. 0.31 ± 1.1). The difference was not significant (P = 0.06).
Children in the intervention group decreased energy intake, while children in the control group slightly increased it; however, the difference was not significant (kcal kg−1 per day; –39.2 ± 89.4 vs. 6.8 ± 55.4, P < 0.06).
Kramer (28–30) (2001)
(2007)
(2009)
Belarus
[2]
PROBIT: Promotion of Breastfeeding Intervention Trial
Cluster RCT
31 maternity hospitals and associated polyclinics (16 intervention, 15 controls)
Sample size: 17,046 mother-infant pairs at baseline (I: 8865, C: 8181). 13,889 were seen at 6.5 years of age (I: 7108, C: 6781).
Newborn full-term singleton infants weighing at least 2500 g with healthy mothers who intended to breastfeed.
Ethnicity not specified.
The experimental intervention was modelled on the Baby Friendly Hospital Initiative. Participants, usually the chief obstetrician and chief paediatrician, from each of the intervention maternity hospitals and polyclinics, respectively, received the 18-h BFHI lactation management training course. The course emphasized methods to maintain lactation and promote exclusive and prolonged breastfeeding.
Duration of intervention: 12–16 months.
Follow-up: at 6.5 years of age of children.
Number of components: 1.
Height, leg length, BMI, head, waist, hip. mid-thigh and mid-upper arm circumference, triceps and subscapular skin-fold thickness.At 6.5 years, no significant effects of the intervention were observed on height, BMI, waist or hip circumference, triceps or subscapular skin-fold thickness. The only slight significant difference was observed for mid-thigh circumference, being higher in the experimental than in the control group: cluster-adjusted difference in means 0.9, 95% CI 0.2–1.5.
Lagström (21) (1997)
Saarilehto (22) (2003)
Talvia (23) (2004)
Hakanen (24) (2006)
Lagström (25) (2008)
Finland
[1]
STRIP: Special Turku Coronary Risk Factor Intervention Project
RCT
Newborn children
Sample size: 1062 children (I: 540, C: 522) from 1054 families at baseline. At 13 years of age, 541 children (I and C not clear).
Age at baseline: 7 months.
Ethnicity not specified.
Families of children in the intervention group received individualized counselling (by a paediatrician, a nutritionist and a nurse) at 1- to 3-month intervals until the age of 2 years, biannually until the age of 7 years, and annually until the age of 13, to reduce the child intake of saturated fat and cholesterol while ensuring their total energy intake. After the age of 7 years, children received additional separate counselling.
Duration of intervention: 13 years.
Follow-up: 13 years and ongoing since start of the intervention.
Number of components (in 0- to 5-year-old children): 1.
Weight, relative weight (as deviation of weight in percentage from the mean weight of healthy Finnish children of the same height and sex), % overweight (defined as a relative weight of +20% or more), BMI at age 8, 10 and 13.Despite finding continuously fewer overweight girls in the intervention group than in the control group after the age of 2 years in the multivariate model, the intervention had no statistically significant effect on the examined growth parameters (at the age of 10 years, 10.2% of the girls in the intervention group and 18.8% of the girls in the control group were overweight; P = 0.0439).
From 4 to 10 years of age, the intervention children received two to three calorie percentage units less saturated fats and 0.5 calorie percentage units more polyunsaturated fats than the control children (P < 0.001).
Mo-Suwan (31) (1998)
Thailand
[0]
Cluster RCT
10 second-year classes of two private kindergartens (four in preschool 1 and six in preschool 2).
Sample size: 310 children at baseline (I: 158, C: 152), 292 children at end of intervention (I: 147, C: 145).
Mean age 4.5 years (SD 0.4).
Ethnicity not specified.
School-based aerobic exercise programme: 15-min walk before the morning class and 20-min aerobic dance session three times per week, by specifically trained personnel.
Duration of intervention 29.6 weeks.
Follow-up: at end of intervention.
Number of components: 1
BMI, triceps skin-fold thickness, WHCU (ratio of weight in kg divided by height cubed in meters).At the end of the intervention, the reduction of the prevalence of obesity in the exercise group was greater than in the control group, although the difference in reduction was not significant (P = 0.057). Girls and boys responded differently to the intervention. Prevention of BMI gain could be shown only in girls. Girls in the intervention group were less likely than those in the control group to have an increasing BMI slope (10.8% vs. 28.1%; OR 0.32; 95% CI, 0.18, 0.56).
Reilly (26,27) (2003)
(2006)
Scotland
[2]
MAGIC: Movement and Activity Glasgow Intervention in Children
Cluster RCT
36 nurseries (I: 18, C: 18)
single blinded
Sample size: 545 children in their preschool year (I: 268, C: 277) at baseline, 504 at 12-month follow-up (I: 245, C: 259).
Mean age 4.2 years (SD 0.2) at baseline.
Ethnicity not specified.
Enhanced PA programme in nursery (three 30-min sessions per week over 24 weeks) plus home-based health education aimed at increasing PA through play and reducing sedentary behaviour.
Duration of intervention: 24 weeks.
Follow-up: at 6 and 12 months after start of intervention.
Number of components: 2.
BMI, expressed as a standard deviation score relative to UK 1990 reference data.The intervention did not have any significant effect on increasing PA, reducing sedentary behaviour or reducing BMI at 6 and 12 months. Mean (SD) BMI z score for intervention at baseline vs. at follow-up: 0.39 (0.98) vs. 0.41 (1.05).

All 17 articles on which this systematic review is based were published in peer review journals and are in the English language. All RCTs were approved by ethics review committees and no conflict of interest was declared. Of the RCTs, three were conducted in North America (two in the USA and one in the USA and Canada) (15–20), two in western Europe (Finland and Scotland) (21–27), one in eastern Europe (Belarus) (28–30) and one in Asia (Thailand) (31). With the exception of the studies conducted in Thailand and Finland, all other RCTs were started after the year 2000. All RCTs but two were cluster randomized trials, with clusters being either preschools/nurseries or school classes in four studies (15–19,26,27,31), and maternity hospitals in one study (28–30). The remaining two studies randomized mother/child dyads (20) and children at the age of 7 months (21–25).

The sample sizes of the RCTs varied considerably, from 43 children in the study by Harvey-Berino et al. (20) to 17 046 newborn infants in the study by Kramer et al. (28–30). Of the seven trials, only two had power calculations on the outcome of interest for our study (19,26), while two had power calculations on the primary outcome of the trial, which was not an anthropometric measure of overweight or obesity (28,32).

Age at enrolment ranged from birth to 36 months. As for the duration of the intervention, five trials lasted 10–30 weeks (15–19,26,27,31), one went on for 12–16 months (28–30) and one lasted 13 years with interventions adjusted to the age of the child/adolescent (21–25). In two trials, the final measurements were carried out at the end of the intervention (16 and 29.6 weeks) (20,22); in one trial, the follow-up time was 12 weeks after the 10.3 weeks' intervention (15), one had a follow-up of 1 year since the start of the intervention (24 weeks) (26,27) and another had a follow-up of 2 years after intervention (14 weeks) (16–19). In the Belarus trial, the intervention was carried out before birth and children were measured at 6.5 years of age (28–30), while the Finnish trial has been going on for 13 years since the birth of children enrolled and the last measurement was completed at the same age (21–25).

Four of the seven interventions were carried out in preschool settings. Of these, one had an exclusive educational component (health, nutrition and PA trainings and behavioural change interventions only) (15), two had an exclusive PA component (excluding education on PA) targeting children (26,27,31) and one included both (16–19). Three of the preschool-based interventions had also an educational module for parents (15–19,26,27). Of the remaining three trials, two were family-based: one study targeted the mother (with an educational approach and parenting support) (20), the second offered individual counselling to both parents and children (21–25). The last intervention was carried out in maternity hospitals-associated clinics and consisted of training obstetricians and paediatricians on breastfeeding (28–30).

Having divided them by type and level of intervention, we present the studies moving from health education only interventions targeting the mother before childbirth, to studies focusing on families and preschools, with a stronger focus on actual PA components.

Educational, training, behavioural change interventions only

Kramer et al. describe a cluster RCT on breastfeeding promotion carried out in the Republic of Belarus (28–30). The Promotion of Breastfeeding Intervention Trial (PROBIT) intervention consisted of the implementation of the 10 steps of the United Nations Children's Fund's Baby Friendly Hospital Initiative, with an extension of the 10th step to include polyclinics in addition to maternity hospitals. For each maternity hospital and associated polyclinic, a chief obstetrician and a chief paediatrician, respectively, received 18 h of training in lactation management. The full implementation of the intervention with the training of all midwifes, nurses and physicians providing care to mothers and children during labour, delivery and postpartum stay, and performing monthly routine visits, required 12–16 months. Recruitment of children during postpartum stay was carried out between June 1996 and December 1997. The intervention on mothers was not described but it is obviously the spill-over effect of the training of health personnel. Overweight and obesity were not mentioned as primary or secondary outcomes of the study in the initial papers (28,33). This, however, is the largest study of the effects of breastfeeding promotion on measures of adiposity in children. Thirty-one maternity hospitals and associated polyclinics were randomized to the intervention or the control group. Out of 17 046 children enrolled, 13 889 (81.5%) were followed up to age 6.5 year. Despite a significant positive impact of the intervention on breastfeeding (28), there was no difference in anthropometric indices at 6.5 years of age, i.e. in the proportion of children with BMI ≥85th and ≥95th centiles, or mean difference in mid-upper arm circumference, triceps skin-fold thickness and subscapular skin-fold thickness (29,30). Additionally, irrespective of intervention, there was no difference reported at 6, 9 and 12 months in weight-for-length for children exclusively breastfed for 3 months or 6 months or more.

The intervention described by Harvey-Berino and Rourke consisted of 11 lessons to native-American overweight (BMI > 25) mothers of 9- to 36-month-old children who could walk. The lessons comprised an educational parenting support programme emphasizing mutual respect, child's encouragement techniques, and psychological and behavioural goals (20). The programme consisted of a parenting support module for the control group and an additional obesity prevention component for mothers in the intervention group. Measurements were taken at start and end of the intervention. Some 43 mothers/child pairs were recruited for the study; three were lost to follow-up during the intervention. The study has several limitations including the small sample size, the relatively short duration of the intervention, and the lack of follow-up after the end of the intervention. Weight-for-height decreased in the intervention group while it increased in the control group (P = 0.06). The authors suggest that this weight change was likely the consequence of a reduction in calorie intake in children in the obesity prevention group. Notably, at the end of the intervention mothers in the intervention group engaged in less restrictive child feeding practices when compared with mothers in the control group (P < 0.05).

In several articles, Lagström et al. describe a 13-year-long prospective RCT started with the enrolment, between February 1990 and June 1992, of 1062 children aged 7 months (21–25). The intervention, part of the atherosclerosis-prevention trial Special Turku (Finland) Coronary Risk Factor Intervention Project for Children, consisted of individualized diet and PA counselling given by a nutritionist and a physician at 1- to 3-month intervals until the child was 2 years old and biannually thereafter. After the age of 7.5 years, children received direct counselling with progressively more information according to age and development of cognitive ability. The families in the control group were seen by the counselling team twice per year until the child was 7 years old and annually thereafter, and only received the basic health education routinely given at Finnish well-child clinics. At 13 years of age, 541 children were monitored (51%) and in a repeated-measure analysis of variance the intervention had no effect on any of the examined growth parameters. At 10 years of age, 585 children were monitored (55%) and the intervention group had not significantly higher proportion of overweight, defined as weight-for-height >20% above the mean on the Finnish growth charts. However, in the univariate analysis, the authors report that at 10 years of age fewer girls in the intervention group were overweight when compared with the girls in the control group (10.2% vs. 18.8%, respectively, P = 0.044). For the boys, the difference was not significant (10.6% vs. 12.1%). The trial, however, had a positive impact on diet with children in the intervention group having – at 4, 7 and 10 years of age – a significantly lower intake of calories per day, a lower calorie-adjusted fat intake and a higher ratio of unsaturated-saturated fat intake.

Dennison et al. describe part of the Brocodile the Crocodile cluster RCT, which consisted in 39 weekly 1-h education sessions in eight out of 16 preschool and day-care settings to promote healthy eating and reduction of television viewing (15). The total duration of the trial was 39 weeks over one calendar year; however, the seven sessions devoted to a reduction in television viewing described in the article were carried out during the second year of implementation of the trial (from fall 2000 to spring 2001) and at least after 10 healthy eating sessions administered during the first year. Each session consisted of a 30-min music activity, 10 min eating a snack and a 20-min interactive educational session. In addition, parents and child care staff were encouraged to implement home- and preschool-based activities (reading stories, turn off the television at mealtime, stimulate children not to watch television for 1 d and 1 week). According to the study flowchart, 93 children in the intervention group and 83 in the control group of mean age 4.0 years were enrolled in the trial in fall 2000, when the sessions on reduction of television viewing started. However, anthropometric data were only available for 102 children, and follow-up data at spring 2001 – end of intervention – were only available for 77 out of the 102 children. The 77 children were comparable in terms of age, height, weight, media use (hours per week), parental education and working status. Children in the study were predominantly Caucasian and from families of middle socioeconomic status. None of the growth measures (BMI for age, standardized BMI for age, triceps skin-fold thickness for age) showed any difference between the intervention and control group throughout the entire study period (measures taken in fall 2000, winter 2000, spring 2001). At the end of the trial, however, children in the intervention group had a significant reduction in mean weekly television/video viewing if compared with children in the control group. No detailed information is available regarding why in fall 2000 complete data were only available for 102 children of the 176 in the flowchart. Furthermore, authors fail to specify how many healthy eating sessions were carried out before the seven evaluated sessions and whether the seven sessions were offered consecutively in 7 weeks or in a longer time span.

Physical activity (only and multi-component)

Mo-suwan et al. carried out an intervention in five of 10 randomized classes of two private kindergartens (31). The intervention exclusively targeted children and comprised just a PA component, consisting of a 15-min walk before morning classes and a 20-min aerobic dance session three times per week led by trained personnel. The intervention lasted 29.6 weeks and measurements (age, weight, height, triceps skin-fold thickness) were made at baseline and at the end of intervention. The activities scheduled for the trial did not interfere with usual preschool activities, including 1 h of physical education per week and, in one of the preschools, an extra 1-h swim class per week. Nothing changed for the children in the control group. Some 310 children were enrolled and analyses were based on 292 children (retention 94%). The change in the prevalence of obesity defined as triceps skin-fold thickness over the 95th centile of National Center for Health Statistics, a division of the U.S. Centers for Disease Control and Prevention (NCHS) charts was not significant neither in the intervention nor in the control group. No significant changes were reported in mean BMI, weight for height cubed (kg m−3) and triceps skin-fold thickness. Significantly fewer girls in the exercise group had an increase in the BMI slope (kg × m−2 × y−1) if compared with girls in the control group. Summerbell et al. describe unpublished data on prevalence of obesity supplied by Mo-suwan et al. and collected 6 months after intervention on 147 children in the exercise and 145 in the control group (34). Even so, differences do not appear to be statistically significant.

Reilly et al. describe the Movement and Activity Glasgow Intervention in Children trial carried out in a randomly selected sample of 36 nurseries with at least 12 children in their preschool year (26,27). This study received the highest score on the Jadad scale among the seven included in our review. Authors emphasized the importance of developing a sustainable intervention (low cost, easily implemented, well accepted by nursery staff and pupils) for increasing PA and improving basic motor skills. Some 545 children were enrolled, of which 504 were followed up for 12 months from the start of the intervention (92%). The intervention comprised activities in nurseries and at home. In nurseries, a 24-week enhanced PA programme was delivered consisting of three sessions of PA each week. To implement the activity, two members of staff from each nursery allocated to the intervention participated in a three-session training. During 6 weeks, intervention nurseries also displayed posters on how to increase PA through walking and playing. Parents of children participating in the intervention received information on how to enhance PA in children and reduce sedentary behaviours such as watching television, with evidence on how low PA was in children from Glasgow. Nurseries in the control group kept their curricula and activities unmodified. Despite an average 71% attendance at prescribed sessions by the children and an 83% of prescribed sessions actually implemented by the nurseries, no significant difference was found between children in the intervention and control groups for PA, sedentary behaviour and BMI.

Hip-Hop to Health Jr. was a 14-week obesity prevention intervention for minority preschool children, which was the focus of a cluster RCT implemented in 24 preschools and described by Fitzgibbon, Stolley et al. (16–19). The preschools involved in the study were part of a programme providing comprehensive education, health, nutrition and parent involvement services to low-income children and their families. Children attending the preschools were predominantly Latinos and African–Americans. Both children and their parents were enrolled in the study, which was divided in two parts. Twelve preschools were first selected and randomized, six to the intervention and six as controls; in the remaining 12 preschools the implementation in the six preschools randomized to the intervention was carried out later. The intervention was monitored from September 1999 to June 2002 in the first group and from fall 2001 to winter 2003 in the second group. The authors present the results on the two parts of the study separately. According to the authors, the intervention was an integrated dietary/PA programme developmentally, linguistically and culturally tailored to minority preschool children, based on social learning and self-determination theories and the stages of change model. The preschool-based intervention consisted of a 45-min session, three times per week for 14 weeks. The 45 min were divided into a short activity to bring the children together, a 20-min interactive educational activity on health eating or exercise and a 20-min aerobic activity. Parents received a weekly newsletter that paralleled the information offered to the children that same week, a weekly homework assignment compensated US$5 when returned, and a twice-per-week 30-min free aerobic class. For 14 weeks, children in the control group were offered a 20-min weekly interactive session on general health activities, such as dental health, immunization, use of safety belt and emergency procedures, while parents received a weekly newsletter on the same general health information children received at school. Baseline, post-intervention and follow-up measurements at 1 and 2 years after intervention were reported. In the first part of the project (retention 73% at 2 years), children were predominantly African–American (89%). No significant difference in BMI was reported in post-intervention measurements between intervention children and controls. At 1- and 2-year follow-ups, mean BMI increase was significantly lower for children in the intervention group when compared with controls. At 2 years, the difference was still significant. The only significant result in dietary intake, PA and sedentary behaviour was, at 1 year, a lower intake of saturated fatty acids in children in the intervention group (% kcal −1.15, 95% confidence interval −1.74 to −0.56). In the second part of the project (retention at 2 years 83%), children were predominantly Latinos and no statistically significant difference in BMI variation was reported at any time point between children in the intervention and control groups. No significant differences were noticed between groups in frequency and intensity of exercise, television viewing, intake of fat and dietary fibres.

Discussion

Our systematic review, which includes more RCTs than previously published systematic reviews and focuses exclusively on children under 5 years of age, shows that to date no single or combined intervention has been able to prove an effect in reducing overweight and obesity, or in limiting weight gain, in preschool children. In some cases, small effects have been observed in proxy variables, such as dietary and/or PA/sedentary behaviours, but none of the examined interventions had an effect on weight gain or BMI. The small effects reported by Mo-suwan in girls (31), by Lagström also in girls and only in the univariate analysis (25) and by Fitzgibbon only in African–American children (18), cannot be considered as evidence of effect.

A recently published systematic review, limited to papers published since 1990 in English, identified and analysed four of the seven RCTs we identified, and reached the same conclusion on the lack of evidence of effectiveness, even though the authors tried to identify effective intervention components on the basis of the impact of the Hip-Hop to Health Jr. trial only on half of the study population (35).

Is the lack of effectiveness found in our and other reviews real, or is it due to faults in the design and conduct of the studies? The failure to show an effect may be due to the choice of the measured outcomes. Most studies have chosen the BMI, or other anthropometric measurements taken at different ages. These may not be the appropriate indicators in children under 5 years of age or slightly older. It is possible that in these children the velocity of growth (BMI centile crossing) (36), or the timing of the adiposity rebound (37), are better indicators of growth exceeding the expected standards, while changes in BMI or other measures of adiposity appear at a later stage. Growth velocity and timing of adiposity rebound, however, would present several practical challenges for data collection, if implemented to detect overweight and obesity in preschool children.

There are certainly faults in the design and conduct of the studies. By examining the Jadad scores assigned to each of the seven trials (Table 2), it appears that the average quality of the studies is low. None of the studies clearly described what happened with the dropouts and virtually none described with sufficient detail the randomization process. In addition, the intervention carried out in Belarus for the promotion of breastfeeding appeared not to have been originally designed to look at overweight and obesity outcomes, but just at breastfeeding rates and duration. Even assuming that the design and the conduction of the studies were good, the interventions involved were not carried out optimally. For example, Campbell et al. (38) suggested that the interventions should have been compulsory, or carried out for longer periods or with longer follow-up. Considering the Hip-Hop to Health Jr. trial, this analysed a very short intervention, targeting both children and parents, and the results (for only half of the sample) were seen at 1- and 2-year follow-up.

Also, the setting of most intervention described in the included studies focussed on the micro-environment (e.g. mothers, families, preschools), while studies on the meso- and macro-environments (neighbourhood, community, society) were lacking. This may be the result of our choice to restrict our review to RCTs. Had we searched for non-RCTs and observational studies, we might have found other types of interventions at a wider community level, but earlier reviews have also noticed that studies on macro environmental factors are largely missing for all age groups (39).

An alternative explanation is that interventions aiming at changing behaviour are mostly ineffective, given the multiple factors involved in behavioural change of which only few are targeted. In fact, in studies where behaviour was an outcome, there was little evidence that behavioural change occurred. It may be that only interventions that aim at modifying also the physical, cultural, economic, social and legislative environment, i.e. the possible upstream drivers of the obesity epidemic, have a chance of actually changing the rates of overweight and obesity. These environmental interventions are not only more difficult to implement, they are also difficult to study because they imply collaborations with agents outside the health sector, e.g. the media or departments in national or local administrations that deal with food prices and availability, traffic, green areas for leisure and PA, food industry, i.e. all those sectors that can bring about changes in the environment that would facilitate changes in lifestyle.

Finally, some of the possible early life determinants of overweight and obesity (maternal diabetes, overweight and smoking, low and high birthweight, rapid infant growth, short sleep duration, consumption of sugar-sweetened beverages, television viewing) (9) have not been included yet in single or composite interventions to be tried out for effectiveness in pregnant women and children under 5 years of age, while the debate on the possible effect of breastfeeding (40) can not be concluded by the results of the PROBIT trial. Research is urgently needed on interventions aimed at changing these determinants and the physical, cultural, economic, social and legislative environment that makes changes in lifestyle difficult.

Conflict of Interest Statement

No conflict of interest was declared.

Acknowledgements

The project this paper is part of was carried out with financial support from the Commission of the European Communities, SP5A-CT-2006-044128 ‘Health-promotion through Obesity Prevention across Europe (HOPE): an integrated analysis to support European health policy’. The study does not necessarily reflect the Commission's views and in no way anticipates the Commission's future policy in this area.

Authors' contribution

  • • Lorenzo Monasta gathered all published and unpublished papers, carried out the analysis and contributed to writing the different drafts of the paper.
  • • David Batty carried out the analysis, and contributed to writing the different drafts of the paper.
  • • Anna Macaluso, Luca Ronfani and Alessandro Bavcar helped searching the databases, contributed to the assessment of the quality of the articles, and to the extraction of the results for inclusion in the tables.
  • • Vittoria Lutje designed the search strategies and searched the databases.
  • • Frank van Lenthe designed and coordinated the HOPE project, and critically reviewed drafts of the paper.
  • • Johannes Brug designed and originated the HOPE project, and critically reviewed drafts of the paper.
  • • Adriano Cattaneo was in charge of the overall project and wrote the different drafts of the paper, including the final version.

Ancillary