Effects of SuperFIT, an overweight-prevention intervention approach, on pre-schoolers’ dietary intake: a pilot study

ABSTRACT SuperFIT aims to improve the dietary and physical activity behaviour of children (2–4 years) in the Dutch childcare and home setting. Healthy parenting practices, policies, and availability of materials, fruits, and vegetables were promoted over the course of a year. This pilot study, with a quasi-experimental design, examined the effectiveness of children’s intake of fruits, vegetables, water, and sweet beverages. Families attending intervention preschools could participate in the partial (preschool) or full (preschool + home) intervention. Parental 24-hour recalls assessed the child’s diet at baseline and two follow-ups. Data from 81 control and 88 intervention children was included in a three-level hierarchical logistic regression. The full intervention showed significantly favourable effects for sweet beverages (B = −1.85, P < 0.05) and unfavourable effects for vegetables (B = −1.94, P < 0.05), compared to the control between the second follow-up and baseline. No significant effects were found for fruit or water, nor for the partial intervention. More research in a larger sample is needed to confirm current findings. Trial registration: ClinicalTrials.gov identifier: NCT03021980.


Introduction
Healthy dietary intake is crucial for a broad range of health outcomes, including obesity (Swinburn et al. 2011;The Netherlands Nutrition Centre n.d.;World Health Organization 2016).Moreover, experiences with food and the formation of food preferences begin at a very young age (Birch, Savage, and Ventura 2007) and tend to track into life-long behaviour (Craigie et al. 2011).Yet, the majority of Dutch young children (1-4 years old) do not meet current fruit and vegetable intake recommendations (150 and 75 grams respectively) (Van Rossum et al. 2016).Furthermore, 62% of the beverages consumed by young Dutch children are sugar-sweetened, making them the largest contributor to young children's sugar intake (Goldbohm et al. 2016).Therefore, young children need to be supported and engaged in the development of healthy dietary habits.
There are several factors that influence young children's dietary intake (Mazarello Paes, Ong, and Lakshman 2015).In order to generate preferences for healthy foods, children need positive and early experiences with healthy foods (Birch, Savage, and Ventura 2007).In fact, repeated taste exposure is considered to promote the acceptance of foods (Ventura and Worobey 2013), and the most successful strategy to increase vegetable intake (e.g.Nekitsing et al. 2018).In addition, children need a supportive environment to be able to develop healthy dietary behaviours.This supports the need to target caregivers, as gatekeepers of these young children's food environment (Birch, Savage, and Ventura 2007).
The majority of young children across OECD countries (Organisation for Economic Co-operation and Development) attend organised childcare (The Organisation for Economic Co-operation and Development (OECD) 2018).In the Netherlands, 55.9% of 0-to 2-year-olds and 88.3% of the 3-year-olds visit an education or care service (The Organisation for Economic Co-operation and Development (OECD) 2018).Childcare has been advocated as an important opportunity to provide fruit and vegetables to young children (Gubbels et al. 2014).In line with this, the use of positive feeding practices by childcare staff is associated with the dietary intake of children (e.g.Gubbels, Gerards, and Kremers 2015).For example, when children were involved in food preparation they ate less sweet snacks, and staff explaining their food preparation activities led to significantly more fruit consumption (Gubbels, Gerards, and Kremers 2015).
Similarly, parents are the gatekeepers of an environment supporting healthy dietary intake at home (e.g.deciding what to eat and what foods are available) and are therefore able to shape their children's dietary intake and food preferences (Kristiansen et al. 2017).Additionally, positive family or parenting practices (e.g.modelling and stimulation of healthy intake) have repeatedly been reported to positively influence children's healthy dietary intake (e.g.Yee, Lwin, and Ho 2017).
These environmental influences emphasise the need to engage both the home and childcare setting in efforts to improve young children's dietary habits.Greater changes in child dietary as well as physical activity (PA) behaviours were seen for children whose parents participated in the nutrition education or at-home activities linked to childcare interventions (e.g. for fruit and vegetable consumption of 2-5 year olds in the U.S. Natale et al. 2014; for vegetable consumption of 3-5 year olds in the U.S. Williams et al. 2014).In addition, a systematic review of various energy balance-related behaviours among 0-4 year olds (van de Kolk et al. 2019) highlighted the importance of direct parental involvement to improve effectiveness.Intervention studies in the childcare setting that included parents merely indirectly (e.g. using newsletters Pinket et al. 2016) or focussed solely on parents disregarding the influence of the childcare setting (Nyberg et al. 2016;Nyberg et al. 2015), generally proved to be less effective.
Considering the above, a comprehensive and integrated intervention approach was developed, targeting both the childcare and home setting directly: SuperFIT (van de Kolk et al. 2020).By creating supportive environments in both settings, SuperFIT aims to improve the dietary and PA behaviour of young children from disadvantaged families in the Netherlands.The present pilot study examines the effects of SuperFIT on the intake of fruits, vegetables, water, and sweet beverages of Dutch children aged 2-4 years.

Study design
The pilot study used a quasi-experimental research design to examine the effectiveness of SuperFIT on children's dietary intake.A study protocol with a detailed description of the SuperFIT intervention and research protocol (van de Kolk et al. 2020), as well as the evaluation of effects on the other primary outcomes (PA and weight status), have previously been published (Van de Kolk et al. 2019).In addition, an elaborate process evaluation has been conducted to assess intervention appreciation, the adoption and implementation of SuperFIT (Harms et al. 2021;van de Kolk et al. 2021).The MUMC + Medical Ethics Committee approved the study.

Setting
SuperFIT was implemented in Dutch preschools.In the Netherlands, 2-4-year olds can attend half-day childcare in preschool which specifically aims to playfully prepare them for primary school (Government of the Netherlands Ministry of Social Affairs and Employment 2019a).The Dutch government stimulates the use of childcare by providing a general childcare benefit based on the parents' working hours and income (Government of the Netherlands Ministry of Social Affairs and Employment 2019b) and provides additional subsidies to children at risk of educational or language disadvantages (Government of the Netherlands Ministry of Social Affairs and Employment 2019c).Preschools therefore reach a broad population (i.e.including underprivileged families).Preschools do not provide main meals (i.e.lunch), as half-day preschool only consists of a three to four-hour attendance per day in the morning or afternoon.Preschools do include one snack moment, at which fruit brought from home is consumed.

SuperFIT
SuperFIT consists of a preschool component and a family component to promote healthy dietary and PA behaviours simultaneously (i.e.promoting fruit, vegetable, and water intake, reducing unhealthy snacking, reducing sedentary behaviour, and promoting active play), given the clustering of these behaviours (Miguel-Berges et al. 2017).Furthermore, SuperFIT is based on systems theories and socio-ecological frameworks (Spence and Lee 2003;Gubbels et al. 2014).
The preschool component of SuperFIT focused on changing the sociocultural (practices of preschool teachers), physical environment (availability of healthy products and play equipment), and political environment (nutrition and PA-related policies) at the preschool (Swinburn, Egger, and Raza 1999).All children at the intervention preschools were exposed to this component.Intervention strategies included training sessions for preschool teachers to increase awareness, knowledge, practices, and skills, related to the target behaviours and environmental factors described above.On-the-job coaching sessions for preschool teachers providing guidance in implementing the key elements of the training sessions (i.e.their practices).In addition, materials were delivered in the form of cards describing hands-on activities (i.e.related to the target behaviours); a weekly distribution of less common fruits and vegetables to increase exposure and variety; a material box with nutrition-related materials (e.g.child-friendly cooking equipment) and play equipment.Parents were not directly involved in this component but were informed through newsletters.
The complementary family component involved parents directly and aimed to promote healthy nutrition and PA at home.Participation was voluntary and therefore only a selection of children in the intervention group was exposed to this component.All caregivers (e.g.parents and grandparents) were allowed to participate in three different types of sessions (caregiver, child, and family sessions) after preschool hours.Caregiver sessions based on Lifestyle Triple P seminars (Bartlett 2017), were organised to increase knowledge about the influence of different environments (Swinburn, Egger, and Raza 1999) on nutrition and PA, discuss barriers they experiences and share experiences with peers.During these sessions, children attended a separate child session in which they played activity games or prepared healthy treats.In addition, family sessions (i.e.caregiver and child together) entailed fun activities to experience healthy nutrition and PA (e.g.making healthy treats and playing simple games to stimulate motor skills).

Study population and procedure
Intervention preschools were recruited through a childcare organisation in Limburg, a province in the south of the Netherlands.Twelve preschools were selected, in accordance with the childcare organisation, based on the low socio-economic status (SES) of the preschool neighbourhood (negative score indicating low SES, calculated based on income, education, and employment status of residents) (The Netherlands Institute for Social Research n.d.).Control preschools were recruited through another childcare organisation in a different region of Limburg, in neighbourhoods with comparable SES scores.There was no geographical overlap between the intervention and control region.This resulted in nine control preschools.All approached intervention and control preschool agreed to participate.No randomisation or blinding was performed to enable organisation-wide policy decisions and support throughout the intervention period.Participating preschools were allocated to the intervention as region-cluster.
All children (2-4 years old) attending the intervention preschools were exposed to the preschool intervention, with the consent of the childcare organisation.Parents were invited to participate in the study, of which 99 (23%) attending intervention preschools and 92 (26%) attending control preschools agreed to.Furthermore, parents of children attending an intervention preschool could participate in the complementary family-based component and additional research concerning this family-based component.This resulted in the full intervention group (n = 47).Parents and children participating in the preschool-based component formed the partial intervention group (n = 52).
The board of childcare organisations provided the consent for inclusion of their preschool in this study.Written informed consent was obtained from the participating child's parents or legal guardian.Participation in the study required at least one parent being able to understand Dutch and both parents signing informed consent.Newsletters and information leaflets were distributed at the preschools, a fun kick-off meeting was organised, and researchers visited all preschools to provide a verbal explanation, and to collect informed consent forms.
A priori sample size calculation was performed, yet recruitment was not able to reach the intended sample size (n = 172 children in each region).An additional power calculation was used to verify the relevance of the obtained sample (n = 191), indicating the obtained sample sizes be able to detect larger effect sizes in the primary outcomes (including dietary intake).A more detailed description of the study population and procedures can be found in the study protocol (van de Kolk et al. 2020).

Measurements
SuperFIT was implemented between April 2017 and May 2018.Baseline measurements (T0) were performed before the start of the intervention, from January to April 2017 for both intervention groups, but continued until July 2017 for the control group due to recruitment difficulties.Two follow-up measurements (T1, T2) were performed in November/December 2017 (after ± 4-11 months) and May/June 2018 (after ± 10-18 months), respectively.

Dietary intake assessment
The present study used an adapted 24-hour dietary recall method, conducted with parents by telephone.Measurements consisted of one phone call each.The dietary recall is a preferred methodology to assess dietary intake due to a lower participant burden and its usability with less literate participants (European Food Safety Authority 2009).Furthermore, parental recall is considered to be a relatively valid and reliable way to assess young children's dietary intake (Livingstone and Robson 2000).
The present study used a dietary recall protocol and software program previously developed for the Good and Healthy Food project (Bessems and Van Assema 2018), based on the multiple-pass approach (Gibson 2005).The protocol was adapted to the intervention aims, focussing on fruit, vegetable, snack, and beverage intake.In addition, adaptions were made based on the target group, such as including child-specific products and quantities (e.g.sippy cup and segments of fruit).
Phone calls were made during evening hours (after dinner time) on Mondays to Thursdays by trained researchers and research assistants.Phone calls were randomly assigned to the researchers, with the exception of parents who felt more comfortable speaking with a researcher in their native language (e.g.Turkish).Calls were not announced beforehand, to prevent an increased awareness of the child's diet (European Food Safety Authority 2009).Phone calls lasted ten to twenty minutes and parents were allowed to decline if the timing was inconvenient.
The protocol required the parent to be present for the greater part of the child's day to prevent missing values.Preschool attendance was not an issue in this respect, as this only entailed three to four hours of a day and one snack moment consisting of standard products (e.g.fruit, water).The protocol structure divided the past 24 h into seven eating moments (EMs): yesterday's evening snack(s) after dinner, today's breakfast, morning snack(s), lunch, afternoon snack(s), dinner, and evening snack(s).
Parents were asked to indicate all food products or meals consumed, starting chronologically with yesterday's evening snack, up to today's last consumption.Interviewers used probing questions (Coulston and Boushey 2008), to ask parents to indicate details about each consumed product.Beverages, fruit, and vegetables were recorded for the main meals (breakfast, lunch, and dinner), and all products consumed in between meals were recorded.Parents were not informed about these specific interests (see Supplementary Table 1 for more details on the recall method).
All products reported were entered in Blaise© (version 4.8.4.1767,Statistics Netherlands (CBS), The Hague, the Netherlands), a system for administering computer-controlled questionnaires.Blaise was integrated with The Dutch Food Composition Database (version 2016/5.0,National Institute for Public Health and the Environment (RIVM), Bilthoven, the Netherlands).Whenever the existing database was insufficient, researchers could use a comment section.

Covariates
A parental questionnaire at baseline was used to collect the child's gender and birthdate (to calculate age) and the parent and partner's birthdate (to calculate age), country of birth, educational level, employment status, and weight and height (to calculate body mass index (BMI)).Questionnaires were completed on paper or online.Country of birth was recoded into 'the Netherlands' or 'other'.Educational level was recoded into 'low'(i.e.primary to lower secondary education), 'medium' (i.e.upper secondary to post-secondary education) and 'high' (i.e.tertiary education and higher) (based on UNESCO Institute for Statistics 2012).Employment status was recoded as 'yes' and 'no'.BMI ≥ 25.0 kg/m 2 was considered overweight.To correct for weather influences (e.g. a heatwave), data on weather conditions was retrieved from the Royal Dutch Meteorological Institute (KNMI) (Koninklijk Nederlands Meteorologisch Instituut 2019).The temperature in degrees Celsius between 6 am and 11 pm was collected to attain an average temperature for each individual recall date.Recall date was recoded into a categorical variable for the season.

Data processing
The Dutch Food Composition Database (version 2016/5.0,National Institute for Public Health and the Environment (RIVM), Bilthoven, the Netherlands) was used to correct dietary intake data for waste (e.g.peels and non-edible parts).The Dutch portion size codebook (RIVM Bilthoven n.d.) provided standard amounts of units or portions (e.g.cups) and was used to recode to grams or millilitres.Multiple corrections were performed, as described below.
First, several unknown and other products were recoded using decision rules (see Supplementary Table 1 for details).Intake levels were cleaned per product per eating moment.In addition, outliers were checked per product per day, using a cut-off value of +4 SD above the mean for each measurement.The sum of intake in grams (solid foods) or millilitres (beverages) was calculated from six eating moments.The last eating moment (EM 7, snack(s) after dinner of the current day) was excluded to cover 24 h.
Fruit, vegetables, water, tea, and all sweet beverages (including regular or diet soda, fruit drinks, and concentrated drinks (including added water), and fruit juice products) were extracted from the dataset.Fresh fruits, dried fruits, frozen fruits, and canned fruit were combined to represent fruit intake.Raw vegetables and prepared vegetables extracted from composite meals represented vegetable intake.Water and tea without sugar (in line with national recommendations) represented the total water intake.All regular soda, diet soda, fruit drinks (excluding 100% fruit juice), and concentrated drinks represented sweet beverages (SB).

Data analysis
Data was analysed using IBM SPSS Statistics for Windows (Version 25.0, IBM Corp, Armonk, NY, USA) and R (Version 3.5.1,R Foundation for Statistical Computing, Vienna, Austria).Frequencies, means and standard deviations were used to describe participants' demographic characteristics at baseline.The effect of SuperFIT was studied with regard to the primary outcomes: vegetable intake, fruit intake, water intake, and SB intake.Explorative effect analyses entailed three steps.For step one, average food intake (g/mL) was summarised through means and standard deviations.Within-group effect sizes (d's) were calculated by dividing the within-group change over time by the standard deviation of the measure in the relevant group at baseline.The second step recoded the daily intake of each separate food group into binary outcomes, as data were not normally distributed (i.e.abundance of 0 g/mL).Participants with a reported intake were categorised as 'yes'; all others were recoded as 'no'.In addition, daily intake of fruits and vegetables were recoded into binary outcomes in relation to the recommend daily amounts (RDA) for children aged 2-4 years, as recommend by the Netherlands Nutrition Centre (Brink et al. 2019).Intakes of ≥150 grams of fruits and ≥75 grams of vegetables per day were categorised as 'yes'; others were categorised as 'no'.In the third step, a three-level hierarchical logistic regression with two random effects (location and subject) to account for the nested structure of the data was conducted to assess the effect of SuperFIT on food intake (yes/no) and meeting the RDA (yes/no).The fixed part of the model consisted of time (T0, T1, T2), condition (control, partial, and full intervention), and the interaction of time*condition to test for intervention effects.Random effect for the location was removed whenever the intraclass correlation was close to zero.All analyses were adjusted for demographic characteristics (child gender, child age in days).Fruit and vegetable intake was adjusted for seasonal variation, known to be of influence (Fowke et al. 2004).Beverage intake was adjusted for average temperature during the day, to correct for heatwaves.Dietary intake was not adjusted for significant differences in parental characteristics at baseline due to missing data.A two-sided p-value ≤ 0.05 was considered to be statistically significant.Similar analyses have been run using linear mixed models.The results were, at least to some extent, in line with the current analyses (data not shown).

Participants
At T0, 191 children and their parents agreed to participate in the study, from which 144 (75.4%) completed dietary intake assessments were available.At T1, this was 107 (56.0%) and at T2, this was 112 (58.6%).Additional analyses revealed no significant differences in demographic characteristics for those who completed data collection at all three-time points compared to those not completing data collection at all three-time points.The present study included all children with at least one completed dietary intake assessment (n = 169, 88.5%).For 116 of these children (66.8%), parents also filled out the parental questionnaire at baseline.Significant differences on participant characteristics at baseline were found for parental birth country (P = 0.006), parental educational level (P = 0.026), and partner being overweight (P = 0.039, Table 1).The control group included more parents who were not born in the Netherlands and who had a lower level of education.3.2.Children's dietary intake

Absolute intake levels
Average daily intake levels are shown in Table 2.The sample size (n = 169) was too small to identify statically significant differences within groups over time, by means of paired ttests.Although standard deviations are rather large, observed levels were explored.The full intervention showed the highest average intake of fruit at T1 and T2, consuming more than the RDA (≥150 grams).The partial intervention and control group consumed on average near the RDA of fruit at all measurements.Within group effect sizes were small.All groups consumed the RDA or more (≥75 grams) at T1 and T2.Both intervention groups showed large increases in average vegetable intake at T1. Within group effect sizes were small.At T0, all groups consumed multiple children's cups of water (150 mL each) on average.The full intervention showed the largest increase and the highest average intake of water at T2.Within group effect sizes showed small, for the partial intervention, to medium effect sizes, for the full intervention and control group, at T2.Finally, the full intervention had the lowest average intake of SB at all measurements.Average SB intake increased for the partial intervention and control group at T2.Within group effect sizes were small.

Any consumption within groups
Proportions of children consuming the food group of interest (≥0 grams or mL) are shown in Figure 1.Again, observed levels were explored.Increasing proportions of children in the full intervention consumed fruit (Figure 1A).The partial intervention, however, showed a decline at T1 and returned to baseline level at T2.For vegetable consumption, increasing proportions were seen for both intervention groups at T1, yet decreased at T2 (Figure 1B).Meanwhile, the control group showed an increasing proportion of children consuming vegetables.For water, ambiguous changes were seen (Figure 1C).An increase was seen for the control group.For SB consumption, the full intervention showed decreasing proportions of children T2 (Figure 1D).Proportions of children remained more or less stable in the partial intervention and increased in the control group.

Meeting the daily recommend amount within groups
Observed levels showed increasing proportions of children that met the RDA (Figure 2).Both intervention groups showed a 'peak' at T1.Still, less than a quarter of the partial intervention group (22.6%) and roughly half of the full intervention group (46.2%) were able to meet the RDA for fruit at T2.Only a minority of the intervention groups (38.5-48.4%)met the RDA for vegetables at T2.The control group showed increasing proportions for vegetable RDA at T2 (Figure 2B).

Adjusted intervention effects
Only two intervention effects were found (Table 3).Probability of consuming any vegetables (>0 g) increased between T0 and T2 in both the partial intervention group and control group, while it decreased in the full intervention group.Probability of consuming    sweet beverages decreased in the full intervention between T2 and T0 while it remained stable in the partial intervention group and control group.

Discussion
The present paper explored the results of the pilot study of SuperFIT regarding fruit, vegetables, water, and sweet beverage intake of 2-4 year olds.Descriptive values showed changes yet limited intervention effectiveness was found.Only children in the full intervention group showed a decrease in consumption of sweet beverages, and an increase in consumption of vegetables, compared to the control group between the second follow-up and baseline.No significant effects were found for fruit or water consumption, meeting the RDA, or for the partial intervention group.Despite various improvements made, changes in the intervention groups were often not large enough to yield significant intervention effects.In addition, the control group improved on several dietary intake outcomes (vegetables and water), causing unfavourable directions in intervention effects.This might be attributable to the control region's participation in other intervention activities.An example is JOGG, an obesity prevention approach for communities based on the French predecessor EPODE (Jongeren op gezond gewicht 2019).Unfortunately, there is no data on specific JOGG activities implemented in the control region, hindering the possibility to differentiate between the effects of SuperFIT and JOGG.
The diversity in results suggests that some intervention activities should be further refined.For example, promoting fruit consumption might be relatively easy to achieve, in light of the natural predisposition to prefer sweet tastes (Mennella and Beauchamp 2010).In contrast, vegetables require children to familiarise with a less sweet taste through repeated exposure (Anzman-Frasca et al. 2012;Caton et al. 2013).Serving barriers (e.g.difficulties with preparing vegetables) hinder pre-schoolers' vegetable consumption at home (Kristiansen et al. 2017).Although intervention activities also addressed providing raw vegetables as snacks, vegetable consumption might rely more on the home setting as they are most often consumed at home (Gubbels et al. 2014).Hence, fruit and vegetable consumption has different barriers and require different intervention strategies (Glasson, Chapman, and James 2011).The additional activities in the full intervention (i.e.caregiver sessions on nutrition and making healthy treats in child and family sessions) might be relatively successful in promoting fruit consumption.For vegetables, however, the intervention activities might have been insufficient in reducing barriers to consumption.Parents may have the motivation to provide vegetables to their child, but lack the physical opportunity (i.e.access to vegetables), social opportunity (i.e.support or social norms) or skills (e.g.cooking skills) to actually do so (Johnson et al. 2018;Glasson, Chapman, and James 2011).Education and family sessions might have been insufficient in providing practical applications or tools to stimulate the integration of elements at home.This might also explain the 'consumption peak' during T1 for both the consumption and meeting the RDA of fruit and vegetables.Fruit and vegetable distribution at preschool was still ongoing during T1, thereby increasing the availability of fruit and vegetables in the childcare setting.Comparable patterns were observed in previous studies assessing the effects of fruit and vegetable distributions, with declines in consumption after distribution stops (Vereecken et al. 2009;Reinaerts et al. 2007).
Increasing the integration of such elements at home could promote sustained effectiveness.
The current study showed the full intervention to be effective in discouraging sweet beverage consumption.Previous research on comparable interventions focussing on healthy beverage consumption showed similar results; with improvements or significant effects for sweet beverages and lack of effectiveness for water (Grummon et al. 2019;Pinket et al. 2016).However, a primary school intervention solely focussing on promoting water consumption through water fountains and educational activities was able to increase water consumption and lacked intervention effects on juice and soft drink consumption (Muckelbauer et al. 2009).Despite the known effects of comprehensive interventions targeting multiple behaviours simultaneously (Ward et al. 2017), it might be the case that caregivers may benefit most from one clear behaviour to change (e.g.replacing sweet beverages with water or consuming more water) as well as an opportunity to perform that action (i.e.water taps).
Parental actions and involvement remain an important element of success.Although mostly non-significant, effects did seem to manifest in the full intervention group only.Similar patterns were seen in a comparable preschool intervention (Lambrinou et al. 2018) and in the effectiveness of SuperFIT on PA (Van de Kolk et al. 2019).Moreover, direct parental involvement in childcare interventions has been reported as a promising strategy to yield child dietary changes (van de Kolk et al. 2019;Hingle et al. 2010).This underpins the need for an integrated system approach, involving both the home and preschool setting (Bradley 2010;Gubbels et al. 2014).To attain optimal alignment among settings, it may be beneficial to consider if and how the SuperFIT family component can be strengthened and further improved to yield greater intervention effects.This is in line with the process evaluation of the family component, which indicated that parents often did not yet actively translate the learned strategies into implementation of these strategies at home (Harms et al. 2021).Alternative strategies to involve parents, e.g. using E-health solutions (Nezami et al. 2018) are perhaps more successful in achieving transfer to the home setting.Future research should look into such different types of parental involvement, beyond direct and indirect involvement.Furthermore, alignment should be optimized between the childcare staff and parents, beyond the settings in general.For example, confident childcare staff might be more willing to discuss healthy dietary behaviours with parents (Willis et al. 2012).

Strengths and limitations
Strengths of the present pilot study included the study design with a baseline and two follow-up measurements; allowing evaluation of both short and long-term effects.Super-FIT also specifically targeted preschools located in low SES neighbourhoods and thus reached underprivileged pre-schoolers.A methodological strength is the use of an adapted 24-hour recall to attain an accurate representation of the children's dietary intake.Each recall measurement was performed for one day only, which was assumed to be sufficient for group mean intake comparisons.However, the day of measurement (e.g.weekend or weekday) was not taken into account, and the dietary recall mostly reflected weekdays.Differences between weekdays and weekends might have influenced the current finding, potentially hiding any associations of the intervention with weekend dietary intake.We therefore recommend future studies assessing intervention effectiveness on dietary intake outcomes to cover all weekdays.In addition, the protocol used has not yet been validated.This is a limitation of the present study.Other methodological limitations include a small study sample.Although the SuperFIT intervention reached a substantial number of pre-schoolers (n = 427), the intended sample size of the present study was not met (van de Kolk et al. 2020).Furthermore, large standard deviations of the outcome variables and prolonged inclusion periods at baseline may have affected the analyses regarding group differences.Interpretation of results should therefore be done with caution, and the findings have to be regarded as explorative.More research is needed to improve the reach and participation in family interventions and related research.
Intervention groups differed on parental characteristics at baseline (see Table 1).However, the analyses examining intervention effects on dietary intake were not adjusted for these differences, as this would have excluded half of the study population due to missing parental characteristics data from the parental questionnaire.This is a limitation of the current study, as parental characteristics are known to influence child dietary intake.It should also be noted that the analyses for fruit and vegetable intake were adjusted for seasonal variation although season was correlated to time in the regression model.In addition, dietary outcomes were not adjusted for total energy intake, as the interviewers focussed solely on SuperFIT's intervention targets: fruit, vegetable, snack, and beverage intake.This might have played a role, as the children grew older as the measurements progressed and might have increased their overall intake levels.
Lastly, there is a lack of blinding due to the quasi-experimental study design and the nature of the study.This increased the chances for the control region to be aware of the research activities.This could have been avoided by randomising preschools.Yet, it is important to note that health promotion programmes are implemented in field settings in which randomisation is usually impossible (Koelen, Vaandrager, and Colomér 2001).It constitutes the dilemma of scientific rigor versus health promotion practice (Koelen, Vaandrager, and Colomér 2001), while balancing the needs of internal validity (i.e.conducting a rigorous study design) and external validity (i.e.generalizability of the intervention approach) (Green 1977).The chosen design allowed for extensive process guidance throughout the entire intervention period.Organisation-wide policy decisions could be made and childcare staff was able to freely discuss their experiences with Super-FIT with colleagues at different preschool locations within their organisation.We believe that this has benefited the support for the implementation process and eventually the impact of SuperFIT.Nonetheless, this also entailed that researchers had a double role, facilitating implementation and gathering evaluation data.

Conclusion
This pilot study regarding the effect of SuperFIT on preschoolers' dietary intake indicated that the full intervention was associated with favourable changes in sweet beverages consumption, but also with unfavourable changes in vegetable consumption.No significant effects on the consumption of fruit, water, or meeting the RDA were found, nor did the children in the partial intervention demonstrate any significant effects.To attain sustainable dietary changes in young children, future interventions should focus on their active integration at home.

Figure 1 .
Figure 1.Proportion of children that, (A) consumed >0 grams of fruit, (B) consumed >0 grams of vegetable, (C) consumed >0 millilitres of water, (D) consumed >0 millilitres of sweet beverage, in each group over time.Significant estimated intervention effects (P < 0.05) compared to the control are indicated with *.

Figure 2 .
Figure2.Proportion of children that, (A) met the recommendations for fruits (≥150 grams), (B) met the recommendations for vegetables (≥75 grams), in each group over time.Abbreviations: RDA: Recommended Daily Amount.Significant estimated intervention effects (P < 0.05) compared to the control are indicated with *.

Table 1 .
Baseline characteristics of participants with at least one dietary intake assessment (n = 169).
Abbreviations: SD, standard deviation. 1 intervention included preschool and family component. 2ial intervention included preschool component only.3Percentagesarebased on available data, N varies due to missing data.4Educationallevel was recoded into 'low'(i.e.primary to lower secondary education), 'medium' (i.e.upper secondary to post-secondary education) and 'high' (i.e.tertiary education and higher) (UNESCO Institute for Statistics 2012).

Table 2 .
Average daily intake of children.
1 Sum of all soda drinks, fruit drinks, concentrated drinks, and fruit juice products

Table 3 .
Hierarchical logistic regression models for all outcomes.