Targeting Sleep, Food, and Activity in Infants for Obesity Prevention: An RCT
OBJECTIVE: The few existing early-life obesity prevention initiatives have concentrated on nutrition and physical activity, with little examination of sleep.
METHODS: This community-based, randomized controlled trial allocated 802 pregnant women (≥16 years, <34 weeks’ gestation) to: control, FAB (food, activity, and breastfeeding), sleep, or combination (both interventions) groups. All groups received standard well-child care. FAB participants received additional support (8 contacts) promoting breastfeeding, healthy eating, and physical activity (antenatal–18 months). Sleep participants received 2 sessions (antenatal, 3 weeks) targeting prevention of sleep problems, as well as a sleep treatment program if requested (6–24 months). Combination participants received both interventions (9 contacts). BMI was measured at 24 months by researchers blinded to group allocation, and secondary outcomes (diet, physical activity, sleep) were assessed by using a questionnaire or accelerometry at multiple time points.
RESULTS: At 2 years, 686 women remained in the study (86%). No significant intervention effect was observed for BMI at 24 months (P = .086), but there was an overall group effect for the prevalence of obesity (P = .027). Exploratory analyses found a protective effect for obesity among those receiving the “sleep intervention” (sleep and combination compared with FAB and control: odds ratio, 0.54 [95% confidence interval, 0.35–0.82]). No effect was observed for the “FAB intervention” (FAB and combination compared with sleep and control: odds ratio, 1.20 [95% confidence interval, 0.80–1.81]).
CONCLUSIONS: A well-developed food and activity intervention did not seem to affect children’s weight status. However, further research on more intensive or longer running sleep interventions is warranted.
- CI —
- confidence interval
- FAB —
- food, activity, and breastfeeding
- OR —
- odds ratio
- POI —
- Prevention of Overweight in Infancy
What’s Known on This Subject:
Obesity prevention in early life has concentrated on changing nutrition and activity in infants, with relatively little success. Although sleep is strongly associated with weight in observational research, few interventions have investigated the effectiveness of sleep modification for obesity prevention.
What This Study Adds:
An intervention targeting food, activity, and breastfeeding did not seem to affect infants’ weight status. Exploratory analyses of the sleep intervention suggest that further research based on more intensive or longer running sleep interventions is warranted.
Rapid increases in childhood obesity,1 and the strong relationship between early rapid growth and subsequent obesity,2 have focused attention on early prevention.3 Although preschool initiatives have shown some success,4 relatively few studies exist in children aged <2 years.5 Three Australian studies have modified some obesity-related behaviors in toddlers and parents,6–8 but only 1 study has significantly affected BMI at 2 years of age.6
Early-life obesity prevention has typically focused on encouraging healthy eating and increasing physical activity, with surprisingly inconsistent results,4,5 thus prompting interest in assessing other behaviors (including sleep).9 Observational studies support a strong inverse association between sleep duration and obesity in childhood,10 and plausible biological mechanisms (eg, changes to eating/activity habits or appetite-regulating hormones) exist to explain this relationship.11 However, whether sleep behavior can change weight trajectories early in life has not been well studied. Existing trials are small and/or target several behaviors as well as sleep12–14 or commence later in infancy.15
Despite the existence of a strong well-child health care system in New Zealand,16 1 in 3 children are overweight or obese by 2 to 4 years of age.17 The aim of the POI (Prevention of Overweight in Infancy) study was to determine whether a conventional approach (food, activity, and breastfeeding intervention [FAB]), and/or an indirect approach (sleep intervention), to obesity prevention would result in lower BMI at 2 years of age compared with standard care.
POI was a 2-year, randomized controlled trial with 4 arms: control (usual care), FAB, sleep, and combination (FAB and sleep) conducted in a single center (Dunedin, New Zealand). Because the protocol is published,18 only essential details are presented here. Ethical approval was obtained from the Lower South Regional Ethics Committee (LRS/08/12/063), and adult participants provided written informed consent. All pregnant women booking into the only birthing unit in Dunedin from May 2009 to November 2010 were eligible if they were aged ≥16 years, <34 weeks’ gestation, able to communicate in English or Te Reo Māori (indigenous language), and planning to live locally for 2 years. Infants were excluded after birth if gestation was <36.5 weeks or they had a congenital abnormality or physical/intellectual disability likely to affect feeding, physical activity, or growth.
Participants were randomly assigned to 1 of 4 study arms, within 6 strata depending on household deprivation (3 levels) and parity (2 levels) by using a block size of 12. Allocation was concealed by using opaque presealed envelopes. Those delivering or receiving the interventions could not be blinded, but all anthropometric assessments were performed by researchers blinded to group allocation, and the biostatistician used uninformative group codes until primary analyses were completed.
Participants in all 4 groups received standard government-funded well-child care (7 core visits from 2–4 weeks to 2 years of age).19 Families in the intervention groups received additional guidance and support (Fig 1). Those in the FAB group received 8 parent contacts,20 including 3 from an international board-certified lactation consultant promoting breastfeeding and delaying the introduction of solids until 6 months.21 Trained researchers (nurses, dietitians, and nutrition graduates) discussed with parents (predominantly mothers) nutrition behaviors believed to affect weight in face-to-face individual sessions at 7, 13, and 18 months of age. The local “Sport Otago” trust held 3 group activity sessions with families to illustrate how to be active with infants and limit time in sedentary activities. Those in the sleep group received a sleep problem prevention program in 2 contacts (antenatal and 3 weeks) regarding developing appropriate sleep habits from birth (trained nurse). Emphasis was on the following: (1) recognizing tired signs and putting the infant down to sleep while awake; (2) without associated settling behaviors (eg, feeding); (3) in a quiet, slightly darkened area; and (4) using safe sleep practices.18,22 Parents who indicated their child’s sleep was a problem from 6 months of age were offered a more intensive personalized intervention, adhering to modified “extinction” techniques.18 Overall, 26.6% of parents in the sleep and combination groups received this extra support. Families in the combination group received the FAB and sleep interventions condensed into 9 contacts (combined antenatal education sessions).
Demographic information obtained at baseline (19–39 weeks’ gestation) included maternal date of birth, ethnicity, parity, education, income, and address (measures household deprivation23). Maternal pre-pregnancy BMI was calculated from self-reported weight at baseline and height measured when the infant was 6 months of age. Anthropometric measures were obtained from hospital data at birth and by trained researchers at 6, 12, 18, and 24 months of age following World Health Organization protocols.24 BMI-for-age z score was calculated by using the World Health Organization growth standards,25 with overweight and obesity defined as the ≥85th and ≥95th percentiles, respectively. BMI and weight status at 24 months were the primary outcomes.
Secondary outcomes were assessed by using face-to-face interviews and telephone questionnaires. Exclusive breastfeeding (no other liquids or solids since birth) and full breastfeeding (no other liquids or solids in the past 48 hours) status to the nearest day were derived from questionnaires administered every 4 weeks from 3 to 27 weeks of age.18 Dietary intake was assessed by using a validated food frequency questionnaire26 at 12 and 24 months.20 Parents indicated how many minutes per week their child spent playing actively outside and inside and watching television at multiple time points. Sleep duration was assessed according to questionnaire responses (parents indicated bed and wake times) and accelerometric findings (24 months only). Children wore ActiCal accelerometers (Philips Respironics, Murrysville, PA) over the right hip 24 hours per day for 5 to 7 days, and sleep duration was estimated by using an automated algorithm.27 Parents reported the number of nights their child typically woke during the week and whether their child’s sleep was a problem (using an 8-point scale). Questions at 12 and 24 months asked the extent to which parents helped their child go to sleep (eg, by touching them, intervening when they woke in the night), who they had received sleep advice from (other than POI), and the usefulness of that advice. Quality of family life28 was measured in mothers and partners when the child was 12 months of age.
The study was designed to have 80% power to detect differences in BMI of 0.5 at 2 years of age, assuming an SD of 1.5, using a 2-sided test at the 0.05 level (n = 142 per arm), with n = 800 in total after allowing for 25% loss to follow-up.18 Analyses were designed to examine questions on the effectiveness of the interventions by using modified intention-to-treat principles (all available data were used for each analysis with participants analyzed as per their assigned group). Self-selected groups (eg, those who chose to receive extra sleep support) are not examined here. Missing data were assumed to be predominantly either missing completely at random or missing at random after conditioning on stratification variables. All analyses adjusted for the stratification variables (ie, 3 levels of household deprivation, 2 levels of parity).
Linear, mixed linear, mixed binary logistic, mixed ordinal logistic, and Cox’s proportional hazards regression were used as indicated in the tables. For outcomes investigated at multiple times, post hoc tests investigated group differences for each time point only if the overall test of group and the group-by-time interaction were statistically significant. If there was evidence of group differences at any particular time point, pairwise group differences were then investigated at that time point. For outcomes with only 2 time points, models for each time point were examined. For continuous outcomes, in which there was evidence of skew or heteroscedasticity in model residuals, natural log transformations were investigated, with a constant of one added if zeros were present. For ordinal logistic regression models, proportionality was examined through comparison with generalized ordinal logistic regression models.
The primary analyses compared the 4 groups with each other (control, FAB, sleep, and combination). We also conducted unplanned (data-driven) exploratory comparisons when the interaction term between the sleep and FAB interventions was not statistically significant (if there was a lack of evidence that the effect of sleep differed depending on whether FAB was present, and similarly for FAB). This approach allowed us to estimate the effects of the “sleep intervention” (ie, sleep and combination compared with control and FAB) and the “FAB intervention” (ie, FAB and combination compared with control and sleep). Details are described in the Appendix.
Stata Release 13 (StataCorp, College Station, TX) was used for all analyses, with 2-sided P values < .05 considered statistically significant. No formal adjustments were made for multiple comparisons. However, to avoid unduly inflating type I error rates, individual tests for outcomes investigated at multiple time points were only performed if there were statistically significant results for prior tests (as described earlier). Nevertheless, marginally significant results should be interpreted with caution.
We recruited 58.1% of those eligible (Fig 2). Participants were older, had less household deprivation, and were more likely to identify as European (all, P < .001) than women who did not consent to participate. Participants were mostly European (85%), 48% were having their first child, and 41% were overweight or obese before pregnancy (Table 1). Retention was high at 2 years, and women who remained were older, less likely to be Māori or Pacific, more highly educated, and from less deprived households (all, P < .05) but did not differ by maternal prepregnancy BMI, parity, or infant sex (all, P > .05). Attendance at intervention sessions was high, particularly during the first year: 95% to 96% of sleep intervention participants received the 2 sleep interventions, and 94% to 96% of FAB intervention participants received the antenatal session and both lactation consultant visits. Seventy-six percent to 90% of participants attended the food and activity sessions in year 1, with 66% attending the final combined session at 18 months.
There was no significant intervention effect observed for BMI or BMI-for-age z score at 24 months (Table 2). Although there was a group difference in the prevalence of obesity (overall, P = .027), this outcome was driven by lower rates of obesity in the sleep group (odds ratio [OR], 0.46 [95% confidence interval (CI), 0.25–0.83]; P = .011) and the combination group (OR, 0.51 [95% CI, 0.28–0.90]; P = .022) compared with the FAB group rather than compared with the control group. There was no evidence of an interaction between the FAB and sleep interventions (P = .755), and an unplanned post hoc comparison showed that the sleep intervention (sleep and combination compared with FAB and control) approximately halved the odds of obesity (OR, 0.54 [95% CI, 0.35–0.82]; P = .004), whereas there was no evidence of an effect of the FAB intervention (FAB and combination compared with sleep and control: OR, 1.20 [95% CI, 0.80–1.81]; P = .385). Given chance imbalances between intervention groups, we repeated the analyses adjusting for maternal BMI (both continuous and categorical) and found no meaningful change in these results (results not shown). As a further sensitivity analysis, in response to the adverse direction of the FAB group (ie, its significantly higher risk of obesity compared with the sleep group, reported earlier), we excluded the FAB group entirely from the model to examine whether all those who received the sleep intervention (sleep and combination groups) differed from the control group. This potentially conservative estimate of the sleep effect was diluted and no longer statistically significant (OR, 0.61 [95% CI, 0.37–1.01]; P = .054).
No intervention effect was observed for the duration of exclusive breastfeeding (P = .069) (Table 3). Exploratory analyses of exclusive breastfeeding duration could not be undertaken because there was evidence of a statistically significant interaction between the sleep and FAB groups for at least 1 time point (P = .035). We have previously reported that there were no differences in food and nutrient intake at 24 months.20 There were a few differences in activity-related behaviors: children in the FAB and combination groups spent more time active outside than the control group at 12 months (both, P ≤ .022), and the amount of time spent watching television was lower in the FAB (P = .014) and combination (P < .001) groups compared with the control group at 6 months.
There was no evidence of a difference in sleep duration, frequency of waking at night, or prevalence of sleep problems (all, P ≥ .187) (Table 3). Although parental ratings of the extent to which their infant’s sleep was a problem differed between groups (overall, P = .023), the only difference was at 27 weeks (overall, P = .042) with higher scores in the FAB group compared with the control group (P = .006). A 3-way interaction involving time (P < .001) precluded exploratory analyses of sleep problems. There was also little variation in child care practices related to sleep, with similar numbers in each group reporting they put their child to bed when tired but still awake (P = .537), did not touch or hold their child to help them go to sleep at night (both P ≥ .314), or allowed the child to self-settle without intervention (both, P > .059).
No significant differences in full or any breastfeeding, family quality of life, or the proportion of children watching television were observed (data not shown). Those in the sleep or combination groups were more likely to report receiving sleep advice (P = .001) but not from nonstudy sources (all overall P ≥ .054), nor were there any group differences in the usefulness of that advice (all, P ≥ .154). Although significant overall group effects on weight (P = .037) and length (P = .039) were observed at 24 months, these changes were relatively small (0.4 kg in weight; 0.7 cm in length).
In this large randomized controlled trial, advice and support on sleep, nutrition, and physical activity did not significantly affect BMI or BMI z score at 2 years of age. The unplanned secondary analysis, however, suggested that those who received the sleep intervention (sleep and combination groups) had a lower prevalence of obesity than those who did not (control and FAB groups). There were few differences in behavioral variables that might explain this reduced obesity risk, including no discernible effect on sleep duration, number of awakenings, sleep problems, or targeted sleep behaviors.
Despite a clear association between short sleep duration and increased obesity risk,10 relatively few interventions have been undertaken. Slower infant weight gain was reported after provision of parental education regarding soothing sleep strategies,12,14 and significant differences in sleep duration and BMI were observed in 2- to 5-year-old children after the use of a multifaceted intervention.13 Two of these studies were small; all included other interventions in addition to sleep and were relatively short (6–12 months), and 2 had no follow-up to determine sustainability. A larger sleep-only intervention reported no significant differences in BMI at 6 years of age after use of a brief intervention (1–3 visits) at 7 to 8 months of age.15 It is possible that earlier intervention, such as in POI, is required. Follow-up will determine whether our intriguing, but unplanned, results showing a positive effect on obesity prevalence at age 2 years remain at 5 years of age.
It is difficult to explain how our sleep intervention might influence obesity if not via sleep duration, which did not differ between groups. Although questionnaires are not always in agreement with actigraphic measures of sleep,29 these objective measures also demonstrated no differences in sleep duration. A major focus of the present intervention was to help the infant fall asleep when tired without external aids.30 Although self-regulation is key to developing healthy sleep–wake patterns31 and is related to body weight in children,32 it is difficult to measure by using a questionnaire. Our findings indicate that similar numbers of children were put to bed while awake but tired across all groups, with few parents touching their child to encourage sleep or intervening before the child self-settled if they woke in the night. Stronger measures of sleep-related self-regulation include objective assessment via video camera, which was not possible in our large trial. However, several well-validated measures of overall self-regulation33 were included in our follow-up measurements, which might determine associations between early self-regulation, sleep, and growth.
We observed reduced television viewing and greater outside active play in the FAB and combination groups, but these differences were small and transient in nature. This lack of effect on nutrition and activity behaviors corresponds with other studies demonstrating relatively few behavioral changes and no effect on BMI,7,8 with 1 exception.6 This latter study was undertaken in a very disadvantaged area of Sydney, Australia, whereas participants in the other 3 studies were predominantly well-educated, European women. It is possible that POI participants were receiving sufficient lifestyle advice from their well-child care, which diminished the potential to alter behavior. However, the high prevalence of overweight and obesity in young New Zealand children (29.6% at 2–4 years of age)17 illustrates the urgent need for additional assistance over and above well-child care. Why it appears to be so difficult to influence nutrition20 or activity34 behaviors at this age is uncertain, but other health priorities such as infant crying35 and sleep problems36 may take precedence for parents. It is also well recognized that parents underestimate their child’s weight status,37 particularly during infancy,38 a time when weight gain is often equated with good parenting.39 Such misperception decreases the likelihood of effective behavior change in response to advice on diet and activity.37
The strengths of our study include high retention and attendance at intervention sessions, as well as collection of outcome data by measurers blinded to participant group. Our study also has some limitations. One-quarter of participants did not complete the 24-month questionnaires, and some demographic differences were observed between those retained and not retained at 2 years. Questionnaires are less accurate at assessing physical activity and sleep behaviors than other methods but were necessary for pragmatic reasons. It is difficult to explain why the sleep intervention affected obesity but did not significantly affect mean BMI. However, both obesity and BMI are important outcomes for successful population-level obesity prevention when mean BMI z scores are well above zero,25 as they are in New Zealand, and our data suggest that the patterns for BMI were similar to those for obesity. Finally, because our sample was relatively socioeconomically advantaged, the findings may be less applicable to those living in more disadvantaged circumstances.
The present nutrition and activity intervention did not seem to affect weight status in children at 2 years of age. Exploratory analyses of the brief sleep intervention (2 face-to-face contacts, with additional support if requested) suggest that further research into more intensive or longer sleep interventions is justified. Future research should also evaluate the potential for sleep to affect growth in groups at higher risk of obesity than was observed in our well-educated, predominantly New Zealand European population.
Appendix: Exploratory Analyses
The primary analyses compared the 4 groups with each other. One of the study’s 4 groups (the combination group) combined both the sleep intervention and the FAB intervention; the study design can therefore also be seen as a 2-by-2 factorial design with an interaction. In effect, we had 2 groups being offered the “FAB intervention” (FAB and combination) and 2 groups being offered the “sleep intervention” (sleep and combination), doubling the number of participants we could observe in response to each intervention and thus providing greater statistical power to detect any effects as long as there was no interaction between the 2 interventions. An interaction would mean that having both interventions simultaneously (ie, being in the combination group) affected the response to the individual approaches, and thus it would not be appropriate to treat the combination group as simply receiving the sum of the FAB intervention and the sleep intervention. The interaction term between the sleep intervention and the FAB intervention was used to determine whether there was evidence that the responses to these 2 interventions were not, in fact, independent. The exploratory (data-driven) comparisons were performed by rerunning the model without the interaction included (ie, a 2-by-2 factorial design without interaction), with variables examining differences between the groups receiving and not receiving the sleep intervention (sleep and combination) and between those receiving and not receiving the FAB intervention (FAB and combination) included in the models (Fig 2).
- Accepted December 13, 2016.
- Address correspondence to Rachael Taylor, PhD, Department of Medicine, University of Otago, PO Box 56, Dunedin 9054, New Zealand. E-mail:
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
FUNDING: Funded by the Health Research Council of New Zealand (grant 08/374) and the Southern District Health Board. Dr Taylor is supported by the KPS Fellowship in Early Childhood Obesity, and Dr Cameron was supported by the University of Otago, Health Sciences Postdoctoral Fellowship.
POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.
- Weng SF,
- Redsell SA,
- Swift JA,
- Yang M,
- Glazebrook CP
- Wen LM,
- Baur LA,
- Simpson JM,
- Rissel C,
- Wardle K,
- Flood VM
- Campbell KJ,
- Lioret S,
- McNaughton SA, et al
- Wake M,
- Price A,
- Clifford S,
- Ukoumunne OC,
- Hiscock H
- Ministry of Health
- Ministry of Health
- Ministry of Health
- Fangupo LJ,
- Heath AL,
- Williams SM, et al
- Cameron SL,
- Heath AL,
- Gray AR, et al
- Salmond C,
- Crampton P,
- Atkinson J
- de Onis M,
- Onyango AW,
- Van den Broeck J,
- Chumlea WC,
- Martorell R
- Hoffman L,
- Marquis J,
- Poston D,
- Summers JA,
- Turnbull AP
- Korkman M,
- Kirk U,
- Kemp S
- Laraway KA,
- Birch LL,
- Shaffer ML,
- Paul IM
- Copyright © 2017 by the American Academy of Pediatrics