Associations between restrictive feeding practices and children ’ s dietary intake: Systematic review and meta-analyses

Background: Parents ’ attempt to limit or restrict children ’ s intake of ‘unhealthy ’ or discretionary foods has been widely considered as a counterproductive feeding practice associated with poorer dietary outcomes, but empirical evidence is varied. Aim: The present systematic literature review aimed to investigate the association between parental restriction and children ’ s dietary intake. Method: Studies were identified through PsycInfo, MEDLINE, CINAHL, Web of Science, and Scopus databases on April 29th , 2022. Included were peer-reviewed, English-language articles published between 2001 and 2022, with an effect size between restriction and children ’ s intake of foods that are ‘healthy ’ (i.e., fruit, vegetables, other general healthy foods) or ‘discretionary ’ (i.e., sweet or savoury energy-dense/nutrient poor foods, high-sugar foods, high-salt/fat foods, and high-energy/sugar-sweetened beverages), or overall diet quality. Risk of bias was assessed using a quality assessment checklist designed to evaluate survey studies. Results: Included studies ( n = 44) were most often conducted in the USA, cross-sectional, and participants were mothers. Effect sizes ( k = 59) from 21 studies were used in nine meta-analyses investigating various healthy and discretionary dietary intake variables. No meta-analytic effects were statistically significant. Qualitative synthesis of effect sizes ineligible for meta-analysis ( k = 91) identified patterns of associations between restriction and increased intake of healthy foods, and decreased intake of discretionary foods. Conclusions: Studies used a diverse selection of measures of restriction and dietary intake, limiting the ability of this review to make accurate cross-study comparisons. However, results suggest that instead of restriction being detrimental for children ’ s dietary outcomes, it may be unrelated, or associated with more beneficial dietary outcomes. Research that utilises validated measures of restriction and dietary outcomes and a longitudinal design is needed to clarify this association.


Introduction
Across developed countries, children's diets are falling short of recommended guidelines for optimal physical and cognitive health (e.g., Australian Bureau of Statistics [ABS], 2018; U.S. Department of Agriculture, 2020; Williams et al., 2020).In Australia, just 6% of children eat the recommended amount of fruits and vegetables (ABS, 2018), while 39.4% of their daily energy intake comes from nutrient-poor, energy-dense 'discretionary' foods (Mitchell Institute, 2019).These inadequate dietary patterns can lead to health issues such as nutrient deficiencies (National Health and Medical Research Council [NHMRC], 2013) and obesity (Craigie et al., 2011), the latter of which increases risk for adulthood chronic illnesses like type 2 diabetes (NHMRC, 2013) and cardiovascular disease (Craigie et al., 2011).As such, there is a pressing need to improve children's diets, which first requires identifying the factors associated with dietary intake.
The behaviours and strategies parents use in feeding their children, known as child feeding practices, have a major influence over children's dietary intake (Russell et al., 2018;Scaglioni et al., 2018).Restrictive feeding practices are of particular interest, as they specifically target dietary intake through limiting the amount and type of foods children eat (Ruzicka et al., 2021), particularly discretionary items high in sugar, salt and/or fat (Fisher & Birch, 1999b).While parents likely adopt a restrictive feeding style to promote their child's health (Musher-Eizenman & Holub, 2007) or to avoid negative social and physical outcomes associated with overweight (Costanzo & Woody, 1985), research has suggested that restriction may be counterproductive.Evidence from early experimental studies indicated that restriction increases children's desire for restricted foods, leading to their greater consumption once made freely available (e.g., Fisher & Birch, 1999a;1999b).Subsequent research by others has replicated this finding (e.g., Jansen et al., 2007Jansen et al., , 2008)).However, most research in this area has used a correlational design and examined parental use of restrictive feeding practices in everyday life and thus has relied heavily on parent-report and often measured restrictive feeding using the Child Feeding Questionnaire's (CFQ; Birch et al., 2001) eight-item restriction subscale.
A portion of the research literature on restriction has demonstrated an association between restriction and higher children's weight status (Ruzicka et al., 2021), although there have been contrary findings indicating a negative prospective association between restrictive feeding and child BMI Z-score (Campbell et al., 2010).However, investigating weight status alone cannot provide information about whether children are receiving adequate nutrition for good health (NHMRC, 2013).As such, researchers have also examined how restriction is associated with children's dietary intake variables.Findings appear highly inconsistent.Some studies suggest restriction is associated with negative dietary intake outcomes, such as lower intake of fruits and vegetables (e.g., Peters et al., 2013), and greater intake of discretionary foods like sweets (Do et al., 2015) and junk foods (Entin et al., 2014).Others suggest restriction is beneficial, encouraging greater intake of fruit and vegetables (e.g., Durão et al., 2015;Wang et al., 2013), and lower intake of discretionary items like energy-dense snacks (Durão et al., 2015) and sugar-sweetened beverages (van der Horst et al., 2006).Other studies have failed to find any statistically significant associations (e.g., Afonso et al., 2020;Campbell et al., 2006;Couch et al., 2014).
To our knowledge no systematic literature review has exclusively addressed the relationship between restriction and children's dietary intake, however four systematic reviews (i.e., Blaine et al., 2017;Faith et al., 2004;Ventura & Birch, 2008;Yee et al., 2017) have examined multiple child feeding practices and child outcomes including dietary intake, eating behaviours and weight status.Faith et al. (2004) investigated relationships between child feeding practices and children's overall energy intake and weight status, and found that restriction was the only practice associated with higher energy intake and weight status.In another review by Ventura and Birch (2008) that examined the relationship between child feeding practices and child eating and weight status, restriction was positively associated with weight gain and problematic child eating behaviours, namely eating in the absence of hunger.Blaine et al. (2017) explored food parenting (i.e., feeding practices) and child snacking, concluding that restriction was associated with greater unhealthy snack intake.In these reviews the most common restriction measure used was the CFQ (Birch et al., 2001), with fewer studies using measures assessing different operationalisations of restriction (e.g., overt and covert restriction; Jansen et al., 2014).Yee et al. (2017) also reviewed the literature on associations between a range of feeding practices and children's (up to age 18 years) intake of 'healthy' and 'unhealthy' foods.The evidence on the association between restriction and intake of healthy and unhealthy foods was mixed.Results of meta-analysis indicated an overall null effect for healthy foods and a small negative effect for unhealthy foods.Notably there was no detail provided on which measures of restriction were used and dietary intake was only considered according to a healthy/unhealthy dichotomy.Taking a more selective approach, Loth (2016) conducted a narrative review examining both pressure-to-eat and restrictive feeding practices and children's dietary intake.It was concluded that while experimental and longitudinal studies suggest a positive association between restriction and intake of discretionary foods, cross-sectional findings are less conclusive.However, the non-systematic, selective nature of this review means conclusions were drawn without considering all available evidence.Across these past reviews (i.e., Blaine et al., 2017;Faith et al., 2004;Loth, 2016;Ventura & Birch, 2008;Yee et al., 2017), there was little or no consideration of how emerging conceptualisations and operationalisations of restriction may influence the association with children's dietary intake.Ogden et al. (2006) and others (e.g., Jackson et al., 2021) have argued that the most widely-used measure, the CFQ (Birch et al., 2001), does not account for the nuances of restrictive feeding practices and includes items which reflect using food as a reward rather than restriction.To address this, Ogden et al. (2006) and later Jansen et al. (2014), distinguish between overt and covert control/restriction in their respective feeding practices questionnaires.'Overt' restriction involves limiting foods in a way a child can detect, and 'covert' restriction involves managing a child's environment to control intake in a way undetectable to the child (e.g., by not bringing unhealthy foods into the home; Ogden et al., 2006;Jansen et al., 2014).Another example of a more nuanced operationalisation of restriction is in the Comprehensive Feeding Practices Questionnaire (CFPQ; Musher-Eizenman & Holub, 2007) where two restriction subscales reflect different motivations for restriction: for weight or for health.These newer conceptualisations of restriction are less researched, and some findings suggest that covert restriction, for instance, is more beneficial for children's dietary intake than overt restriction (Jansen et al., 2014;Loth, 2016;Rollins et al., 2016).There is therefore a need for a current systematic review to comprehensively assess how restriction, across its many and varied conceptualisations, is associated with children's dietary intake.
The present paper aims to systematically review the literature investigating the association between different operationalisations of restrictive feeding practices of parents of children aged 2-12 and children's dietary intake.We focus on three broad categories of dietary intake: intake of 'healthy' foods such as fruits and vegetables, 'discretionary' foods with low nutritional value and which may be high in sugar/salt/fat, and global measures of overall diet quality based on intake of a wide range of foods.As findings in this field are mixed and therefore inconclusive, this review aims to provide some clarity by synthesising the available research through both qualitative and metaanalytic methods.

Search strategy
This systematic literature review and meta-analyses followed PRISMA guidelines (Page et al., 2021).PsycInfo, MEDLINE, CINAHL, Web of Science, and Scopus databases were searched on 29th April 2022.Titles, abstracts, and database-specific thesauruses were searched, using terms developed for the concepts of restrictive feeding practices, children, and dietary intake (see supplementary materials, Table S1, for search terms).Search results were limited to peer-reviewed, English-language articles, published from 2001 to 2022.Earliest publication year was limited to the CFQ's development in 2001 (Birch et al., 2001), as it is one of the earliest and most widely-used measures of restriction (Rollins et al., 2016).

Inclusion and exclusion criteria
Articles were included if they: a) were peer-reviewed, b) were published in English, c) were published from 2001 to 2022, d) included a sample of parents of a child aged 2-12 years, e) included one or more survey measure of restrictive feeding practices (i.e., restriction of discretionary/unhealthy foods or drinks), f) included a measure of children's intake of 'healthy' foods (i.e., fruits, vegetables, general healthy foods such as healthy snacks and 'core', 'traditional' or 'wholesome' foods; Johnson et al., 2011;Xiang et al., 2021;Ystrom et al., 2012) and/or 'discretionary' foods (i.e., sweet or savoury energy-dense/nutrient-poor foods, high-sugar foods, high-salt/fat foods, high-energy/sugar-sweetened beverages), or overall diet quality, and g) reported an effect size for the relationship between restrictive feeding practices and children's dietary intake.Children's age was limited to older than two years to exclude articles measuring breastfeeding or complementary feeding.If the sample included a proportion of parents whose child was aged below two or above twelve years, the study was included if this applied to less than 10% of the sample.Longitudinal and randomised controlled trial studies were included if they provided baseline/comparison group data; for studies commencing before children were two years old, the first published article when children reached the 2-12 age range was included.
Articles were excluded if: a) they were a review or purely qualitative paper, b) children were not within the 2-12 age range (or >10% were less than 2 or older than 12 years old), c) measure/s were child-reported, d) variables were measured at different time-points (i.e., >6 months), e) the measure of child feeding practices did not include a clear, exclusive measure of restriction (e.g., Parental Feeding Styles Questionnaire [Wardle et al., 2002] and the Caregiver's Feeding Styles Questionnaire [Hughes et al., 2005]) or did not focus on the restriction of discretionary/unhealthy food (e.g., restriction of main meals or restriction of fruit juice/drinks), and g) dietary intake was only measured in terms of proportion of energy intake from fat, or total energy intake (e.g., total kilocalories consumed), as this does not provide adequate information about children's diet quality.Studies assessing eating in the absence of hunger using the paradigm developed by Fisher and Birch (1999a) rather than children's usual dietary intake were also excluded.

Screening and extraction
Screening and extraction were conducted by LW in consultation with KM.Duplicate articles were removed using EndNote version 20.2 (EndNote 20), with further duplicates removed through Covidence (https://www.covidence.org).Titles and abstracts were screened against inclusion and exclusion criteria, with remaining articles screened at the full-text level.Information extracted from eligible articles included author, year of publication, study design, sample size, gender and age of the children and parent sample, measure or measures used to assess restrictive feeding practices (i.e., restriction of discretionary/unhealthy foods or drinks) and children's dietary intake, and effect sizes for the association between restriction and dietary intake.For this review, bivariate Pearson's correlations were the most appropriate effect size for potential meta-analyses.When studies only reported standardised regression coefficients (beta coefficients), authors were contacted for zero-order correlations.Of the 23 authors contacted, two provided correlations (i.e., Burnett et al., 2021;Haszard et al., 2015), five authors responded but no longer had access tho the data, 16 did not respond (of these two authors' email addresses were no longer active).

Quality assessment
Study quality was assessed using the Quality Assessment Checklist for Survey Studies in Psychology (Q-SSP; Protogerou & Hagger, 2020).Twenty items assessing different article domains were assigned "Yes", "No", "Not stated clearly", or "Not applicable" responses.Overall study quality score was calculated as the percentage of "Yes" answers within total applicable items.Depending on amount of applicable items, scores of at least 70-75% indicated "acceptable" quality, while scores below these percentages indicated "questionable" quality.Question 4 was adjusted so operational definitions of variables could be provided in the methods section.Question 11 was adapted so that requirements were met if a study cited a measure, rather than provided it in full.Lastly, questions 18 and 19 were adapted so that if an ethics committee approval was listed, this was taken as evidence that informed consent/debriefing occurred.

Data synthesis and analysis
Studies were eligible for meta-analysis if they reported an appropriate effect size (i.e., Pearson's correlation or standardised regression coefficient) for the relationship between restrictive feeding practices and children's dietary intake.Three or more effect sizes involving the same restriction measure and dietary intake variable were required for each meta-analysis.If studies reported separate effect sizes for groups within their sample (e.g., urban/rural location), these were entered into meta-analyses individually.Lastly, studies using validated measures were required to use at least half of their original items to be eligible for meta-analysis (e.g., at least four items from the eight-item CFQ Restriction subscale; Birch et al., 2001).
Meta-analyses were conducted using the Meta-Essentials 'Correlational data 1.5' workbook (Suurmond et al., 2017), using random-effects models with Fisher's r-Z-r transformation (Fisher, 1921).Pearson's correlations were the preferred effect size metric (van Rhee et al., 2015).However, due to Peterson and Brown's (2005) suggestion that imputing missing Pearson's correlations with beta coefficients provides reasonably accurate effect size estimates, beta coefficients were used when Pearson's correlations were unavailable.Their inclusion makes use of relevant effect sizes, potentially reducing sampling error (Peterson & Brown, 2005).If only unstandardised regression coefficients were available, beta coefficients were calculated using the standard deviation of dependent and independent variables when available.Effect sizes not suitable for the planned meta-analyses were summarised qualitatively.Meta-analytic effect size estimates and correlations included in qualitative synthesis were interpretated based on Lovakov and Agadullina's (2021) guidelines: 0.12 was considered small, 0.24 medium, and 0.41 large.

Publication bias
Publication bias was suggested by statistically significant Egger regression (Egger et al., 1997) and Kendall's Tau a (Begg & Mazumdar, 1994) values.Funnel plots of standard error by effect size were inspected visually for each meta-analysis, with values distributed symmetrically on either side of the combined overall effect size suggesting no publication bias.

Results
Fig. 1 presents a flow diagram of the article inclusion process.Six full-text articles appeared to meet all inclusion criteria, but were excluded because they used the same sample and similar measures as another eligible article with an earlier publication date.These were Boots et al. (2017Boots et al. ( , 2018aBoots et al. ( , 2018b)), Durão et al. (2017), Papaioannou et al. (2013), andSleddens et al. (2014).Of the 44 studies included in the review, 23 were included in meta-analyses.Eight of these studies also included effect sizes ineligible for meta-analysis, which were synthesised qualitatively (i.e., Boots et al. (2015); Burnett et al. (2021); Gubbels et al. (2009); Haszard et al., 2015;Jansen et al. (2021); Jarman et al. (2015); Taylor et al., 2017;Tysoe & Wilson, 2010).A further 21 studies were included in qualitative synthesis only, as they lacked effect sizes eligible for the planned meta-analysis.

Study characteristics
Table 1 presents a summary of key study characteristics and quality assessments for included articles.Studies were most commonly conducted in North America (n = 13) and Australia (n = 8).Most studies used a cross-sectional design (n = 27), with longitudinal designs also relatively common (n = 12).Overall, mothers were over-represented.Thirty studies included both mothers and fathers, yet most often at least 80% of participants were mothers.The remaining 14 studies used mother-only samples.Half (n = 22) of all studies used the CFQ (Birch et al., 2001) as a measure of restrictive feeding practices.Children's dietary intake was most often assessed using a previously-validated or author-developed food frequency questionnaire (n = 23).Also common were 24-h dietary recalls (n = 7), and validated dietary questionnaires (n = 8) such as the Children's Dietary Questionnaire (n = 4; Magarey et al., 2009).

Quality assessment
Using Q-SSP criteria (Protogerou & Hagger, 2020), the quality of most (n = 35) studies was acceptable, with the remaining nine of questionable quality (see supplementary materials, Table S2, for study quality assessments).Frequent issues included no justification/rationale for sample size, no attempts to treat attrition, lack of evidence for validity of measures, and no information about the duration of data collection.

Meta-analyses
Nine meta-analyses assessing the relationship between restriction and children's dietary intake variables were conducted: five for children's 'healthy' dietary intake variables and four for children's discretionary food intake variables (see supplementary materials, Table S3, for studies included in each meta-analysis).
Table 2 presents effect size estimates and number of beta (β) coefficients used in the meta-analyses involving the following 'healthy' dietary intake variables: fruits only, vegetables only, fruits and vegetables (measured together), 'healthy'/core foods (e.g., healthy snacks;   Boots et al., 2015), and overall diet quality (e.g., Burnett et al., 2021).All meta-analytic effect sizes were small and not statistically significant.Table 3 presents effect size estimates and number of beta (β) coefficients used in meta-analyses assessing restrictive feeding practices and the following 'discretionary' dietary intake variables: discretionary foods that can be sweet or savoury (e.g., unhealthy snacks; Boots et al., 2015), savoury (high-salt/fat) foods, sweet (high-sugar) foods, and high-energy/sugar-sweetened beverages.All meta-analytic effect sizes were small and not statistically significant.

Heterogeneity and publication bias
For all meta-analyses, I 2 and Cochran's Q values indicate large (Higgins et al., 2003) and statistically significant (Cochran, 1954) levels of heterogeneity (see Tables 3 and 4).However, due to the small number of studies per meta-analysis, I 2 values may be biased (von Hippel, 2015), and Cochran's Q values may be underpowered (West et al., 2010).
Overall, Kendall's Tau a and Egger regression values suggest publication bias was unlikely (see Tables 2 and 3); all Egger regression values were non-significant.However, these tests may be imprecise for smaller meta-analyses such as those in the present review (Begg & Mazumdar, 1994).Funnel plots of standard error by effect size appeared generally symmetrical (suggesting no publication bias), with some asymmetry in the meta-analysis assessing general healthy foods.

Qualitative synthesis
The qualitative synthesis included effects (k = 91) from 29 of the 44 studies included in this review.Of these, 23 studies were ineligible for meta-analysis, and eight contributed to meta-analysis but also reported effects that were excluded from meta-analysis.

'Healthy' dietary intake variables
Forty effect sizes from 19 studies assessed the relationship between  a 24-h dietary recalls require participants to recall their dietary intake from the preceding 24-h period, usually via a structured interview (Olukotun & Seal, 2015).
These may be 'single', 'double', 'triple', or 'multiple pass', depending on how many stages are used to collect data (e.g. the first stage may include a brief listing of foods consumed, while a subsequent stage may include asking for the time and place foods were eaten; Cambridge Biomedical Research Centre, n.d.).

Table 2
Meta-analytic effect size, heterogeneity, and publication bias estimates for associations between restrictive feeding practices and 'healthy' Children's dietary intake variables.
Children  4, most of the effects were null (k = 29).Of the 11 statistically significant associations, nine were small and positive, reflecting that higher restriction was weakly associated with higher intake of 'healthy' foods.In particular, for overall diet quality three of the four effects were significant and positive.However, there were no significant relationships between restriction and vegetable only intake and only one significant positive effect for fruit only intake.There were a mix of significant positive and negative associations for fruit and vegable intake assessed together.Whilst there was no clear overall pattern found according to restriction measure or sub-type of restriction, four out of five effects that assessed covert restriction (Ogden et al., 2006) and a healthy dietary intake variable were positive and significant (Table 4).

'Discretionary' dietary intake variables
Fifty-one effect sizes across 24 studies assessed the relationship between various operationalisations of restrictive feeding practices and 'discretionary' dietary intake variables (i.e., sweet or savoury discretionary foods, high-sugar foods, high-salt/fat foods, and high-energy/ sugar-sweetened beverages).As shown in Table 5, most effects were non-significant (k = 35).Of the 16 statistically significant associations,

Table 4
Summary of effects (not included in meta-analysis) reported between different operationalisations of restrictive feeding practices and Children's intake of healthy foods and overall diet quality.Note.↑ = statistically significant positive relationship (i.e., higher intake); ↓ = statistically significant negative relationship (i.e., lower intake); ns = non-significant relationship.
most were negative (k = 11), and occurred most commonly with intake of discretionary sweet/savoury foods or sweet (high-sugar) foods.For the high-energy/sugar-sweetened beverages variable, there was a mix of negative and positive associations with restrictive feeding practices.No clear pattern emerged regarding restriction measures/sub-type and discretionary dietary intake variables (Table 5).

Discussion
This systematic review aimed to synthesise through both metaanalytic and qualitative methods the literature investigating the association between restrictive feeding practices and children's intake of healthy foods, discretionary foods and overall diet quality.Overall the review revealed limited evidence to suggest that restriction is associated with either positive or negative outcomes for dietary intake in children 2-12 years.
Effect size estimates produced by all nine meta-analyses were small in magnitude, with none reaching statistical significance.Thus, no evidence was found to suggest a positive or negative relationship between restrictive feeding practices (mostly measured by the CFQ restriction subscale, Birch et al., 2001) and children's intake of 'healthy' or discretionary foods, or overall diet quality.This is contrary to the findings of some previous reviews (i.e., Blaine et al., 2017;Faith et al., 2004;Loth, 2016;Ventura & Birch., 2008), which suggest that restriction is negatively associated with intake of healthy foods, and positively associated with intake of discretionary foods.However, these reviews did Note.↑ = statistically significant positive relationship (i.e., higher intake); ↓ = statistically significant negative relationship (i.e., lower intake); ns = non-significant relationship.(u/w & n/w: underweight and normal weight; o/w: overweight and obesity.a Studies in which parents were asked to report on restriction of specific foods only. not conduct meta-analyses, so their findings rely on purely qualitative syntheses.Our findings are more aligned with the systematic review and meta-analysis of Yee et al. (2017) who found restriction to be associated with lower intake of unhealthy foods; with effect sizes larger for older children.Furthermore, while past reviews (i.e., Blaine et al., 2017;Faith et al., 2004;Loth, 2016;Ventura & Birch., 2008) suggest restriction is associated with poorer dietary intake outcomes, the present qualitative synthesis found mixed results.Among statistically significant results, restriction was most consistently associated with higher intake of healthy foods (e.g., fruit and vegetables) and overall diet quality and lower intake of discretionary foods (both sweet or savoury) and high-sugar foods.Findings were mixed for high-energy/sugar-sweetened beverages intake, with both positive (e.g., Vaughn et al., 2017) and negative (e.g., Mazza et al., 2022) associations reported.Taken together with the null findings of the meta-analyses in the present paper, there is little evidence to support the conclusions of previous reviews (i.e., Blaine et al., 2017;Faith et al., 2004;Loth, 2016;Ventura & Birch., 2008) that restriction negatively impacts children's diet.Potential reasons for the inconsistency of the literature are considered in section 4.1.

Limitations of the literature
The major limitation in the literature was the wide range of restriction measures used, each operationalising restrictive feeding practices differently.For example, while the CFQ (Birch et al., 2001) assesses restriction that is more 'overt' (Rollins et al., 2016), the FPSQ-28 (Jansen et al., 2016) accounts for both overt and covert restriction in separate subscales, possibly leading to very different results.The validity of many restriction measures is also questionable.Many studies develop measures without assessing their validity, while others use shortened versions of existing validated measures, possibly negating their psychometric properties.In the present review there was no clear evidence that the measure of restriction used in a study influenced the results, however given that the majority of studies used the CFQ restriction scale, there were insufficient studies to assess these specific conceptualisations (and operationalisations) of restriction and dietary outcomes meta-analytically.Despite its apparent popularity, the validity of the CFQ restriction subscale has been criticised (for a detailed account refer to Jackson et al., 2021).Further research to investigate different conceptualisations of restrictive feeding and child outcomes is warranted.Another limitation identified in this review was the lack of justification provided in studies for the sample size.Inadequate sample sizes and therefore under-powered studies may be in part responsible for the null findings of the meta-analyses.
Measures of children's dietary intake were similarly diverse, substantially impacting the clarity of cross-study comparisons and ability to conduct meta-analyses.In the studies reviewed children's dietary intake was often measured via a food frequency questionniare which assessed the frequency but not the specific amount of foods eaten.This presents a limitation in terms of the questionable validity of dietary intake measures used across studies.It may also explain inconsistent patterns of association identified in the qualitative analysis between restriction and vegetables only intake, fruit only intake and fruit and vegetable intake combined.

Limitations of the present review
A significant limitation of the present review process was the amount of beta coefficients used in meta-analyses.While beta coefficients were imputed for unavailable Pearson's (r) correlations to utilise available effect sizes and reduce sampling bias (Peterson & Brown, 2005), these coefficients were sourced from analyses involving other independent variables (e.g., parent education level or child age).This potentially impacted the precision of, and likely underestimated, meta-analytic effect sizes (Roth et al., 2018).Thus, the over-reliance on beta coefficients may partially explain the null findings across meta-analyses.
Another limitation of this review was that parent-reported measures of children's dietary intake were considered.While this criterion was chosen to promote consistency in diet measurement across studies, parents cannot indicate with certainty what their child has eaten in their absence (Rollins et al., 2016).This may help to explain why a positive, rather than negative, association was found between restriction and beneficial children's dietary outcomes.Rollins et al. (2016) suggests restriction positively influences children's dietary intake while parents are present, yet leads to increased discretionary food intake when children are unobserved.Therefore, this review may have captured data regarding intake when parents are present and can implement restrictive practices.
Lastly, several studies did not report statistically non-significant findings for the relationship between restriction and children's dietary intake.Consequently, individual effect sizes (e.g., Kröller & Warschburger, 2008;Tysoe & Wilson, 2010) or entire studies (e.g., Jang et al., 2019;Vollmer et al., 2015) were not included in the review, reducing the available evidence for both meta-analyses and qualitative synthesis.

Future research
Future research should be consistent in its measurement of both restriction and children's dietary intake, to allow for more robust comparisons of results.Specifically, as overt and covert restriction may differentially influence children's dietary intake (Loth, 2016;Rollins et al., 2016;Vaughn et al., 2017), more research should assess these concepts using validated measures like the FPSQ-28 (Jansen et al., 2016).For greatest accuracy, children's dietary intake should be assessed through multiple pass 24-h recalls (Burrows et al., 2010).For older children, self-report of dietary intake, especially of foods eaten away from home (e.g., at school), is a viable option (Yee Lai & Eiin Wong, 2023) and avoids the limitations of parental social desirability bias.Additionally, more longitudinal and experimental research should be conducted to investigate the relationship between restriction and children's dietary intake over time.These study designs will allow researchers to better identify causality or temporality in these relationships than cross-sectional studies (Blaine et al., 2017;Loth, 2016).

Conclusion
This systematic literature review did not find conclusive evidence for a relationship between restrictive feeding practices and children's dietary intake.Results of meta-analyses indicated no significant association, however qualitative synthesis of remaining effects suggested some evidence that restriction may be beneficial for children's diets.These was no clear differentiation between different operationalisations of restriction (e.g., overt and covert) and children's dietary outcomes based on the limited research available.As the outcomes of the present review are overall contrary to the general consensus in the literature, it is essential more research is conducted that is consistent in its measurement of both restriction and children's dietary intake.Hopefully, this will provide evidence to make informed recommendations regarding the use of restrictive feeding practices, ultimately improving children's diets and health.

Ethical statement
"Ethics approval was not required for this systematic review as only data from published articles was used."

Fig. 1 .
Fig. 1.Flow diagram of the article screening and inclusion process.

Table 1
Summary of key characteristics and quality assessment ratings for studies included in the systematic review (n = 44).
A(continued on next page) L.M.Werner and K.M. Mallan
= Higgins et al.'s (2003)ts for analysis; k = number of independent effect sizes; r = weighted correlation coefficient; CI = confidence interval.I 2= Higgins et al.'s (2003)heterogeneity measure; Q = Cochran's (1954) heterogeneity measure.Positive r values indicate greater intake of the dietary variable of interest.Discretionary foods (both sweet and savoury) include energy-dense/nutrient-poor foods categories that can be either sweet or savoury.For example, unhealthy snacks (e.g., Brown

Table 5
Summary of effects (not included in meta-analysis)reported between different operationalisations of restrictive feeding practices and Children's intake of discretionary foods.