Parent–child correlation in energy and macronutrient intakes: A meta‐analysis and systematic review

Abstract In the current study, we aimed to review the evidence from twin and family‐based studies that have assessed the familial similarity in intakes of energy and macronutrients among various parent–child pairs. The online literature databases, including Web of Science, PubMed, and Scopus, were searched up to December 2022 to find potentially eligible studies. We converted Pearson's, Spearman's, or intra‐class correlation coefficients to z's using Fisher's z transformation to obtain approximate normality and then calculated a mean and standard error (SE) of transformed correlation weighted by the sample sizes in the studies. We reported pooled r and 95% CI as our final results in five groups, including parent–child, mother–daughter, mother–son, father–daughter, and father–son. Twenty‐one eligible studies were included in this meta‐analysis, in which the sample size ranged from 33 and 4310. Our analysis showed that family resemblance in the intake of energy and macronutrients in various parent–offspring pairs was weak to moderate which could be different based on family pairs, nutrients, and studies. The highest similarity in dietary intakes was observed among the mother–daughter pair, which was for carbohydrate and protein intake, respectively. The lowest correlations in dietary intakes were found between mother–son or father–son pairs. Our meta‐analysis suggested that family similarity for intakes of energy and macronutrients was not strong in parent–child pairs. The highest correlation in dietary intake was mostly found in mother–daughter pairs. The weak similarities in dietary intake among parent–child pairs indicate the noticeable effect of the environment outside the family on individuals' dietary choices.


| INTRODUC TI ON
Children's eating habits are influenced by both genetic and environmental factors, and parents can have an impact on both of these factors (Savage et al., 2007).They play a pivotal role as both gatekeepers and role models in influencing their children's nutrition beliefs and behaviors, which can have a significant impact on their offspring's dietary patterns and long-term health outcomes (Beydoun & Wang, 2009).Hence, it is commonly believed that children's dietary intakes are strongly linked to their parents, owing to the combined influence of genetic and environmental factors on eating habits (Laskarzewski et al., 1980;Rossow & Rise, 1994).However, there is controversy regarding the results of studies on familial resemblance in dietary patterns.
While some studies provide evidence supporting familial resemblance in dietary patterns (Laskarzewski et al., 1980;Oliveria et al., 1992;Pérusse et al., 1988;Rossow & Rise, 1994), other studies have found that the association is either very weak or nonsignificant (Feunekes et al., 1997;Feunekes et al., 1998;Lahmann et al., 2017).This is likely because of the multifactorial nature of people's dietary patterns, with the family being responsible for only a part of it (Popkin, 2006).For example, as children age, their independence in food choices increases, and during this time, the influence of their peers on food choices becomes more apparent (Bogl et al., 2020;Nicklas et al., 2004).Thus, changes in children's dietary intake over time can also have an impact on the parentchild dietary association.
Considering these factors, it is important to conduct a systematic examination and quantification of the association between parent-child dietary intakes.Therefore, the main objective of this study is to conduct a systematic review and pairwise meta-analysis of studies published since 1980 and evaluate the extent of correlation and similarity between parent-child dietary intakes.In addition, we performed meta-regression analysis to compare correlations based on their type of parent-child relationship, dietary assessment approach, sample size, and other potential sources of heterogeneity.
The results of this study are expected to contribute to our knowledge of the factors that influence dietary patterns among young people, which is important for developing effective public health interventions and policies aimed at promoting healthy eating habits and reducing the risk of chronic diseases.

| Systematic search
This systematic review and meta-analysis was performed according to the preferred reporting items for systematic reviews and meta-analyses (PRISMA) guidelines (Figure S1).Three major databases including PubMed, Scopus, and Web of Science were searched up to December 2022.Also, a manual search was performed in the Google Scholar database and reference list of found papers to find potential eligible studies.Two reviewers independently screened studies first by title and abstract and at the next level by full-text reviewing.

| Eligibility criteria
Each observational study that reports a correlation between energy and macronutrient intake among family member pairs is included in our study.There was no limitation for the age of participants and publication date.Familial relationships that were required for our study were: parent-child (PC), mother-child (MCH), father-child (FCH), father-son (FS), father-daughter (FD), mother-son (MS), and mother-daughter (MD).
Nevertheless, studies that weren't in English, didn't report sample size, reported r after implementing an intervention, or were conducted on unhealthy populations were excluded.In a few studies that did not announce the sample size for each group in the final analysis, but mentioned values were available at baseline, we calculated the approximate final sample size (e.g., we summed the parent's population with the child population to obtain parent-child sample size).

| Data extraction
To have the minimum number of missing papers, two reviewers independently extracted the required data, shared their results, and resolved any inconsistencies by review.First author name, year of publication, region of studies, type of relationships, age of participants, the correlation coefficient presented by r (even intraclass, Pearson, or Spearman rank correlation coefficients) and its related sample size, dietary assessment method (either FFQ, dietary records or recalls) and unit of measurement was extracted from included studies.

| Quality assessment
Newcastle-Ottawa quality assessment scale for cross-sectional studies was used to evaluate the studies' quality (Table S1).It measures studies' quality on 3 levels and checks them for selecting a representative sample, outcome and exposure assessment, comparability and cofounders' control, and using a proper statistical test.
Eventually, a score of ≥7 out of 10 indicates the good quality of the studies.

| Statistical analysis
The familial correlations of dietary intake of energy, carbohydrate, protein, and fat were evaluated during this study, and pooled r and 95% CI were reported for each of them in the following groups: (1) father-son, (2) father-daughter, (3) mother-son, (4) mother-daughter and (5) all (including mentioned groups besides to parent-child, father-child, and mother-child).
We converted Pearson's, Spearman's, or intra-class correlation coefficients to z's using Fisher's z transformation to obtain approximate normality and then calculated a mean and standard error (SE) of transformed correlation weighted by the sample sizes in the studies (Donner & Rosner, 1980).Due to the number of studies and their conceptual heterogeneities, we run the random effect method for every analysis.I 2 statistic was used to present this heterogeneity and p heterogeneity <.05 was considered significant (Wang et al., 2011).
As different studies report different units of nutrients for their measurements (e.g., calories, joules, gram, serving, percent of energy, or proportional terms like g/kg), in this case, we analyzed them in three levels.First, we pooled all correlation coefficients from every unit.
If a study reports both gram and percent correlation coefficients for one variable, we analyzed them at the same time point.In the second and third levels, we analyzed studies that report the percent of energy and gram of intake, respectively.Evaluation of publication bias was conducted using Egger's test, Begg's test, and visual inception of funnel plots.Regarding the number of included studies and to simplify our final results, Egger's test values were considered for reporting.
Regarding analyzing different nutrients in different pairs, and discrepancies in studies' population characteristics, there was conceptual and objective heterogeneity of reported correlations; so, we performed an unadjusted model of meta-regression analysis to find some variables that could be a potential source of heterogeneity.In this order, we used Z and SE Z statistics and reported the beta coefficients, 95% CI, p-value, and tau 2 (τ 2 ) for each variable.τ 2 statistic indicates between-study variance and is an estimate of heterogeneity in the results.Therefore, if τ 2 is reduced by adjusting the effect of a variable, then the heterogeneity is reduced and that variable can be considered as a possible source of heterogeneity.Therefore, meta-regression was done based on the year of publication (published paper after 2000 vs. before 2000), children relationships (parent's correlation with girls vs. boys), parent relationships (children correlation with mothers vs. fathers), coexistence (parentchild correlations that not living together vs. living together), region (Asia, Oceania, and Africa vs. Europe and America), dietary assessment method (dietary records or recalls vs. FFQ), child's age (older than 18 years old vs. younger), sample size (larger than 500 people vs. lower than 500 people), and variable unit (percent of energy vs. gram of intake).Decision on studies that report age range was based on the population of different age groups or total mean age calculation using each group's mean age and its sample size.
Finally, we tested differences between the mean correlation of energy and macronutrient intake for each familial pair (FS, FD, MS, MG) using analysis of variance (ANOVA) and pairwise t-test to find nutrients that significantly had different correlations in a family pair.We also examined differences among familial pairs for each nutrient correlation using ANOVA and pairwise t-test to find family pairs who had the highest or lowest correlation coefficient of each nutrient.For all analyses, the significant levels were considered as p < .05.Statistical analysis was done using MedCalc software (version 20.218, Ostend, Belgium; https:// www.medca lc.org; 2023) and STATA software version 17.0.

| RE SULTS
3.1 | The literature searches 1892 publications were initially identified from databases.After excluding 398 duplicates and 1426, non-relevant papers, 68 full-text papers of potentially relevant studies were detected.Of 68 articles, 53 articles were eliminated based on the inclusion and exclusion criteria and 15 papers remained.In addition, 6 articles were obtained from other sources, and finally, 21 articles were included in this meta-analysis (Figure 1).
Table 3 indicates the pooled results of the meta-analysis on macronutrient correlation in family members.Also, the range of correlations for macronutrient intake, pooled r (95% CI), the percent of heterogeneity, and the p-value of the Egger test for evaluating the publication bias are stated in Table 3.The highest pooled r was reported for fat (percent of energy intake) (0.23), total CHO (percent of energy intake) (0.22), protein (percent of energy intake) (0.22) among FS, protein (percent of energy intake) (0.27), fat (percent of energy intake) (0.20), simple CHO (percent of energy, grams of intake, and servings/day) (0.19) among FD, protein (percent of energy intake) (0.25), total CHO (percent of energy intake) (0.23), and total CHO (percent of energy, grams of intake, and servings/day) (0.22) among MS, and total CHO (percent of energy intake) (0.27), protein (percent of energy intake) (0.26), and fat (percent of energy, grams of intake, and servings/day) (0.23) among MD.The highest pooled r among all members of the family was reported for protein (percent of energy intake) (0.25), total CHO (percent of energy intake) (0.24), and fat (percent of energy intake) (0.21).
Regarding the diversity of included articles in essential variables, there is significant heterogeneity in the meta-analysis on the correlation of macronutrient intake in family members except for total CHO (percent of energy, grams of intake, and servings/day) among MS, Simple CHO (percent of energy, grams of intake, and servings/ day) among MD, protein (percent of energy intake) among MD and MS, protein (grams of intake) among FS.We evaluated the publication bias using the p-value of the Egger test, no significant publication bias was observed in our meta-analysis.
The meta-regression on energy and macronutrient intakes to people vs lower than 500 people), and variable unites (percent of energy vs gram of intake) among family members is shown in Table 4.

| Meta-regression for family members
There is an inverse relationship between year (coefficient: −.09; 95% The difference in mean r between family members paired with energy and macronutrient intake is reported in Table 5.The findings indicated a significant difference in mean r between MS and MD in energy intake (p-value <.05).However, there was no significant difference in the mean r among other family members paired for intakes of energy and macronutrients.
Table 6 shows the mean r differences in family member pairs including FS, FD, MS, and MD according to energy and macronutrient intakes.There are no significant differences in paired groups consisting of energy and macronutrient intakes among all paired family members.b Since this study did not announce the sample size by groups, the sample size for each group was obtained from the sum of the values mentioned at the beginning of the study. c The studied population is divided into two groups, including family members who live together (subgroup 1) and who live separately (subgroup 2).

TA B L E 1 (Continued)
In this meta-analysis study, we focused on the results of various studies that had investigated the possible family similarity in dietary intakes (energy and macronutrients) between various parent-child pairs including (parent-child, father-son, father-daughter, mother-son, and mother-daughter).The pooled r based on studies results showed that family correlation of dietary intakes between various parent-child pairs can be varied according to types of parent-child dyad, nutrients, and studies.The correlation reported for dietary intakes, including energy, carbohydrate, fat, and protein was weak to moderate based on different parent-child pairs.The highest family similarity in dietary intakes was found among mother-daughter pairs, which included carbohydrates and proteins, respectively.
The lowest family correlations in energy and macronutrient intakes were mainly found between mother-son or father-son pairs.Various factors, including the year of publication of the paper, the dietary assessment approach, the coexistence of parent-child, the child gender, and child age were the sources of heterogeneity in the results of this meta-analysis.The interesting results of this meta-analysis provided insight into the extent of the parental effect on offspring food habits and dietary intakes and showed that there was a weakto-moderate correlation in intakes of energy and macronutrients among various parent-child dyads.The weak to moderate family similarity in dietary intakes among parent-child pairs suggested that although dietary intakes of individuals may somewhat be influenced by genetic affinities between the parents and offspring, the noticeable effect of shared and non-shared environmental factors should not be missed.Previous studies that reported weak correlations for dietary intakes between parent-child pairs suggested that dietary behaviors of offspring, especially in adolescence and youth periods, can be influenced by various environmental factors in addition to household factors, including the influence of friends and peers, the effect of food environments of community, workplace, and school, the influence of the type of environment and programs used to spend leisure time such as television viewing, as well as personal factors, such as self-esteem, autonomy, and self-image (Beydoun & Wang, 2009;Boynton-Jarrett et al., 2003;French et al., 2001;Salvy et al., 2007;Satia et al., 2001;Shariff & Yasin, 2005).
This meta-analysis of the findings of previous studies shows that family similarities in energy and macronutrient intake in mother-daughter pairs are somewhat stronger than in other parent-child pairs.Although the reasons for these findings are not fully known, some factors can justify this interesting finding (Hosseini-Esfahani et al., 2022;Park et al., 2004;Wang et al., 2009); in most societies, mothers and daughters spend more time during the day than fathers and sons in the home environment and may have a greater joint role in the process of cooking and preparing food for the family's main meals and hence their food choices may be more similar in meals throughout a day (Hosseini-Esfahani et al., 2022).However, sons may spend more time outside the home environment with peers and schoolmates, and under the influence of these people, they may have different food choices than their family and parents (Hosseini-Esfahani et al., 2022;Park et al., 2004).So, this point has caused the correlation of dietary intake between parents-son, especially father-son, has been reported to be weak in previous studies.Also, boys and girls potentially have different biological, psychosocial, behavioral, and physiological differences, which can explain the difference in the effect of genetic and environmental interaction on family similarity reported for mother-daughter pairs compared to other parent-child pairs (Wang et al., 2009).Regarding that, based on biological differences, the period of physical growth, maturity, and the stage of body development of sons and daughters occurred at different times of life, therefore, the nutritional needs of sons and daughters may be differed according to the influence of these factors that subsequently cause differences in dietary behaviors between them during adolescence and the later stages of life.For instance, at a certain age during adolescence, girls have fully matured physically, but boys are still growing and maturing physically.
Moreover, physiological differences in sons and daughters may affect differences in similarities of dietary intakes between parents and offspring; because of possible differences in body parts and physical strength, sons may be more physically active than daughters, which may cause them to have different food intake during the day.Generally, according to the points mentioned above, the effects of the home environment, outdoor environments (school, workplace, restaurant, etc.), parents, peers, classmates, and friends on sons and daughters can be different, which can lead to differences in dietary intakes and food choices between them.
Results of our meta-analysis revealed that familial similarity in the intake of macronutrients as a percentage of energy is higher compared to the conditions in which the intake of macronutrients for individuals has been determined as grams per day; according to the results of meta-regression, this can be considered as a source of heterogeneity in the results of the studies.The above-mentioned finding supports the claim that the similarity of genetic characteristics in parent-child can cause them to act similarly in terms of allocating a specific share of daily energy intake for different macronutrients (carbohydrates, protein, and fats).Therefore, although the intake of macronutrients by parents and children as g/day may not be very close to each other, it seems that the genes or single nucleotide polymorphisms (SNPs) involved in determining the percentage of macronutrients from total energy in the form of a dietary pattern for parents and children are similar, and therefore, this has caused that the correlations for the intake of macronutrients as a percentage of energy has been observed stronger than their intakes as g/day among parent-child pairs.
Our findings based on a meta-regression showed that the resemblance of dietary intake between older children and their parents was lower than their younger counterparts; some previous investigations have suggested the considerable effects of factors outside the home as reasons for these results (Beydoun & Wang, 2009;Hosseini-Esfahani et al., 2022).As children grow older, they spend more time outside the home and are more likely to be influenced by their peers.
Thus, during adolescence and youth, offspring show more autonomy in their food choices and may behave differently than their parents in dietary intake (Beydoun & Wang, 2009).One of the sources of heterogeneity in the results of the present meta-analysis was the issue of coexistence; we showed that the family similarities in dietary intakes among parent-child not living together versus those living together were different.The overall lower dietary correlations reported in the children not living with their parents revealed that family resemblance and heritability in dietary intakes are not considerable in offspring living apart from their parents and can weaken e Parent-child correlations that not living together versus living together.
f Others (Asia, Oceania, and Africa) versus Europe and America.
g Dietary intakes assessed by records or recalls versus FFQ.
h Children older than 18 years old versus younger.
i Study population larger than 500 people versus lower than 500 people.
j Units of variable (percent of energy versus gram of intake).
TA B L E 4 Meta-regression analysis for energy and macronutrient intake among all family members a .
with time.The influence of other people, such as peers, and also higher autonomy in food choices are factors that cause children who live apart from their parents to make possibly more changes in their food choices than those who live at home with parents (Hosseini-Esfahani et al., 2022;Lahmann et al., 2017;Zuercher et al., 2011).
In the current meta-analysis, the year of the study was a source of heterogeneity in the extracted findings.We showed that studies performed before 2000 have stronger family similarities in macronutrient intake among parent-child pairs than studies conducted after 2000; these results revealed that the diversity in food intakes in individuals in recent years (after 2000) was higher in compared to several decades ago (before 2000) because, in times before 2000, individuals mostly adhered to a traditional dietary pattern that was mainly consumed in the family environment, this is even though in recent years, with the occurrence of nutrition transition in various societies (Popkin et al., 2012), individuals' dietary patterns have undergone extensive alteration and they shifted from the traditional and steady dietary pattern to the consumption of fast foods, which are usually consumed outside the home environment (Oexle et al., 2015;Popkin et al., 2012).Also, the extensive changes and developments in the food industry especially in Western countries, and creating various changes in cooking, and preparation methods of foods caused family members, including parents and child may have different types of food in their daily meal plan, some of which foods may be consumed outside the family environment, such as fastfood restaurants, coffee shops, workplaces, etc. (Oexle et al., 2015;Popkin et al., 2012;Tansey & Worsley, 2014).
Based on our findings, the difference in the dietary intake assessment approach (FFQ vs. food record or 24-hour recall) was another source of heterogeneity in the findings of studies.We indicated that in the studies that used FFQ to obtain nutritional information, the family similarities observed for different parent-child pairs were weaker than in studies that used the 24-hour recall or food record for the assessment of dietary intakes.These results can be justified and accepted; because it has been previously demonstrated that using FFQs for the collection of food intake data has major limitations in comparison to other dietary assessment approaches, such as diet diaries or dietary recalls (Olafsdottir et al., 2006).Indeed, FFQs are not optimal nutritional assessment instruments to determine actual dietary intakes.FFQ may not reflect lifetime dietary habits and only provides a general estimation of dietary intakes, while to assess the absolute intake of energy and macronutrients, using a 24-h recall (for several days) or food diaries may be much more useful and practical (Olafsdottir et al., 2006;Teucher et al., 2007).Thus, FFQ that had been used to collect dietary intake data in several familial studies was a source of heterogeneity (in the form of a negative factor) that has caused a decrease in the reported values of correlation coefficients for energy and macronutrient intakes among parentchild pairs.
The most important application of the results of this metaanalysis for public health can be that the weak-to-moderate similarities in dietary intakes in parent-child pairs show that modification in parents' food habits will have a weak to moderate alteration in children's dietary behaviors; these effects are expected to be seen more in younger children (i.e. in a child less than 10 years old).It is also possible that making dietary interventions in mothers will have a greater impact on improving the food choices of children, especially girls.
Several strengths of this study deserve mention.This metaanalysis is the first to review and summarize the evidence from all previous studies that examined familial similarity in dietary intakes among different parent-child pairs.We also included in this TA B L E 5 Comparison of mean r differences between energy and macronutrient intake among family member pairs a .

TA B L E 6
Comparison of mean r differences between family member pairs based on their energy intake and macronutrients.meta-analysis as much as possible any study that had examined familial similarity for each macronutrient or energy intake in each parent-child pair, so that we could conduct a comprehensive review of study findings for all macronutrients and energy intake.Furthermore, it should be noted the included studies in this meta-analysis have been conducted in different societies which have people with different demographic, socioeconomic, and nutritional characteristics, therefore, the results of the present meta-analysis on the possible family similarities of the intakes of energy and macronutrients may be possibly generalized to different populations.Nevertheless, it is necessary to mention some limitations.Most of the studies on family similarity in dietary intakes among parent-offspring have been conducted in Western countries or developed societies, and studies in developing countries are limited.Considering that lifestyles, nutritional behaviors, and as well as family relationships can be different in these societies, therefore, it is necessary to obtain an accurate summary of the degree of family similarity in dietary intakes, more studies should be performed in developing countries.
Also, the meta-regression analysis in this study showed the possible heterogeneity in the findings of eligible studies conducted on the similarities in intakes of energy and macronutrients among various parent-child pairs.The year of publication of the paper, the dietary assessment approach, the coexistence of parent-child, the child gender, and child age were the sources of heterogeneity in the current study; this heterogeneity in the selected studies may lead to the problem of making a single and accurate conclusion based on extracted findings.Furthermore, Due to the lack of sufficient data in the studies included in the present meta-analysis, we were unable to determine whether children >18 years old living in the same household as the parent or apart from them, how much it affects the parent-child correlation in energy and macronutrient intakes.

| CON CLUS IONS
The present meta-analysis suggested that family resemblance of dietary intakes in various parent-offspring pairs can be different based on family pairs, nutrients, and studies.In general, family similarity for dietary intakes, including energy, carbohydrate, fat, and protein was weak in all different parent-child pairs.The strongest dietary intake associations were observed in motherdaughter pairs in which the highest correlation was for carbohy-

F
find the potential sources of heterogeneity including year (published paper after 2000 vs before 2000), children (parent's correlation with girls vs boys), parent (children correlation with mothers vs fathers), coexistence (parent-child correlations that not living together vs living together), region (Others including Asia, Oceania, and Africa vs Europe and America), dietary assessment (dietary intakes assessed by records or recalls vs FFQ), child age (children older than 18 years old vs younger.),sample size (study population larger than 500 The main characteristics of included studies assessed familial resemblance of macronutrient intake among family members.
drate and protein intake, respectively.The weakest correlations for dietary intakes were mainly shown in mother-son or fatherson pairs.Findings of the current investigation reported that although dietary intakes of people may somewhat be influenced by genetic affinities between the parents and child, the noticeable influence of environmental factors should not be ignored.The weak family similarities of energy and macronutrient intake among various parent-child pairs indicate the effect of environmental factors on individuals' dietary choices, such as the effect of peers, school and classmates, workplace, etc.
Statistical distribution of included studies based on their characteristics in both absolute and proportional terms.Pooled results of correlations of macronutrient intake among family members.
TA B L E 2 Presented as the number of papers/number of studies.Percent of energy, grams of intake, and servings/day.The bold values in the table are statistically significant.a All papers that report energy and macronutrient intake in all familial pairs were included.
a b Fisher's transformed.c Not Significant heterogeneities are bolded.d Significant publication biases are bolded.e f Percent of energy intake.g Grams of intake.c Parent's correlation with girls versus boys.d Children correlation with mothers versus fathers.
FD, father-daughter; FS, father-son; MG, mother-daughter; MS, mother-son.a Significance level considered p-value <0.05 and significant differences are bolded.