|Home | About | Journals | Submit | Contact Us | Français|
We used meta-analytic methods to examine the frequency of shared family mealtimes in relation to nutritional health in children and adolescents. The primary objective was to determine consistency and strength of effects across 17 studies that examined overweight and obese, food consumption and eating patterns, and disordered eating.
The total sample size for all studies was 182 836 children and adolescents (mean sample age: 2.8–17.3 years). Pooled odds ratios were calculated. A random-effects model was used to estimate all outcomes.
The frequency of shared family meals is significantly related to nutritional health in children and adolescents. Children and adolescents who share family meals 3 or more times per week are more likely to be in a normal weight range and have healthier dietary and eating patterns than those who share fewer than 3 family meals together. In addition, they are less likely to engage in disordered eating.
Educational and public health initiatives aimed at promoting shared family mealtimes may improve nutritional health of children and adolescents. Clinicians may advise their patients about the benefits of sharing 3 or more family mealtimes per week; benefits include a reduction in the odds for overweight (12%), eating unhealthy foods (20%), and disordered eating (35%) and an increase in the odds for eating healthy foods (24%).
Study results have suggested that family mealtimes may act as a protective factor for many nutritional health-related problems during childhood and adolescence, including issues of overweight, unhealthy eating, and disordered eating. Findings have been mixed with some studies that reported strong relations to health outcomes such as obesity, whereas others have reported no relation.1,2 These inconsistencies make it difficult to inform parents of the relation between family meals and health outcomes. Inconsistencies may stem from the variability in ages examined and the concentration on both genders rather than examining them separately. Although the American Academy of Pediatrics recommends that families regularly eat meals together as part of its initiative of what families can do to prevent childhood obesity,3 pediatricians could benefit from more precise estimates of the effects of shared family mealtimes on children's nutritional health. In this meta-analysis, we determined the strength of the relationship between the frequency of shared family mealtimes and children's nutritional health.
We were interested in 3 major public health concerns: obesity, unhealthy eating, and disordered eating. In particular, we examined the effects of sharing 3 or more meals per week versus 1 or none. When study design allowed, we investigated the long-term potential for family meals operating as a protective factor for these health indicators.
Four search engines were used to systematically locate empirical research: PubMed, PsycINFO, Web of Science, and the Cochrane Database of Systematic Reviews. We conducted the study in 2009, and there were no year restrictions. The following key words were used: (“family” or “interaction”) and (“mealtime” or “dinnertime”).
Inclusion criteria were that (1) studies must have had a measure of family meal frequency, (2) outcome variable(s) must have included at least 1 of the following categories: obesity, food consumption (ie, [un]healthy eating), and disordered eating, (3) the data must have used an estimate that was an odds ratio (OR) or could be transformed into an OR, (4) outcome variables must have been child or adolescent focused, (5) studies must have been peer-reviewed, and (6) studies must have been written in English. Authors were contacted when information was not present in the article. A total of 1747 studies were identified initially, but 1730 were excluded (see Fig 1 for exclusion reasons).
The following information was coded: (1) average age of sample participants; (2) study design (longitudinal or cross-sectional); (3) gender mix of the sample; (4) control for confounding variables; and (5) outcome of interest. For purposes of being thorough, 2 coders independently examined all dimensions (1–5) for all included studies. Agreement was high (89%). In addition, the reference and works-cited lists of all included studies were reviewed to capture all potential studies. All information was entered into the Comprehensive Meta-analysis (CMA 2.0 [Biostat, Englewood, NJ]) program.
Most studies (n = 12) asked participants to consider the number of family members present for the meal. For instance, some asked participants to report on how often some, most, all, or, more broadly, other family members shared a meal together. Other studies just asked participants to report on how often regular family dinners occurred but made no mention of the number of family members present (n = 3). Two studies asked participants to report only on shared meals that had at least 1 parent present.
Three categories were constructed on the basis of reported outcomes: weight status, food consumption, and disordered eating. Consistent with the definition used most commonly across the studies, overweight was defined as having a BMI at the ≥85th percentile.4 Food consumption was grouped into 2 categories to reflect the consumption of unhealthy versus healthy foods. Food consumption was measured by using a food-frequency checklist, and unhealthy foods included soda, fast food, fried food, and sweets/candy. Unhealthy eating also included the absence of healthy foods such as skipping breakfast and not eating at least 2 fruits or vegetables in a day. These foods were usually separated into categories on the basis of total intake (eg, greater than once per day to rarely/never consuming). Healthy eating included fruit and vegetable consumption, multivitamin use, and breakfast consumption. For fruit and vegetable consumption, responses were dichotomized into yes/no responses on the basis of whether the person met the national recommendations per day. Disordered eating included bingeing/purging, extreme and less extreme weight-control behaviors (ie, taking diet pills, self-induced vomiting, using laxatives, using diuretics, fasting, eating very little food, using food substitutes, skipping meals, and smoking cigarettes as a means of weight control).
Each study was used as the unit of analysis. Adjusted ORs with 95% confidence intervals (CIs) were used to estimate the relation between family meal frequency and the outcomes. All studies reported ORs with the exception of 1 study,5 which reported only frequency data; therefore, an OR was calculated on the basis of available information. In the analyses that examined obesity, food consumption (ie, unhealthy eating), and disordered eating, ORs of <1 indicate a protective factor. For healthy eating, ORs of >1 indicate a protective factor. If a study reported multiple outcomes, mean effect sizes were calculated for “independence concerns,” limiting 1 effect size per study. If each outcome is treated as being independent, 2 problems can occur. One issue is that studies with multiple outcomes are weighted more heavily; the other problem involves underestimating the precision and overestimating the error. The ideal way of handling multiple outcomes is to compute the mean of the outcomes for each individual study and then to use this score as the unit of analysis.6 This approach also takes the correlation among the outcomes into account. Furthermore, if a study reported outcomes for boys and girls separately, or younger and older adolescents separately, these outcomes were combined to yield an overall effect size for that study. In addition, if a study shared a sample with another study, the outcomes were combined to contribute only 1 total outcome in the analyses that examined the overall effect of family meal frequency. However, in the analyses that examined separate categories of outcomes, studies that shared the same sample but reported on different outcomes were able to report a unique effect size for each category of outcome. ORs that were reported by using “never having family meals together” as the reference group were transformed into their reciprocals, and corresponding 95% CIs were computed.
The random-effects model was used to calculate pooled effect sizes (ORs) with 95% CIs. We used the Q statistic to test for heterogeneity in the effect sizes. For studies with large samples, a nonsignificant value suggests that the effect size comes from the same population and that the finding is robust across studies. If the value was significant, moderators were examined. We also used I2 because the Q statistic and corresponding P value only address 1 aspect of heterogeneity—that of the true dispersion that equals 0. This statistic, I2, is a degree of inconsistency and measures the proportion of variance that is true rather than random error.6 An I2 value close to 0 means that nearly all of the variance is a result of random error, whereas an I2 value farther away from 0 may reflect real variance and indicate the potential for a moderator analysis. For the situation in which moderators were tested, we only tested for 1 moderator, because it is not recommended to test for more when the number of studies is small (n < 10).6 To assess publication bias we relied on the classic fail-safe N,7 which is used to address the “file-drawer problem.” The fail-safe N yields the number of missing articles that would have to exist to bring the effect size to nonsignificance. The larger the number is, the more stable the finding. Small numbers indicate that the findings should be interpreted with caution.
Table 1 lists descriptive information for the 17 included studies.1,2,5,8–21 The total sample size across studies was 182 836 participants (range: 145–99 462 participants*). The mean sample age ranged from 2.8 to 17.3 years. In terms of geographical makeup, the studies were conducted in the United States (n = 12),1,8–12,14,15,17,19–21 Australia (n = 1),2 Canada (n = 1),13 Finland (n = 1),18 Japan (n = 1),5 and New Zealand (n = 1).16 All studies were adjusted for confounders with the exception of 1 study.5 Twelve studies1,2,8–12,14,17,19–21 controlled for race/ethnicity. Similarly, 12 studies1,2,9–12,14,17–21 reported information on socioeconomic status (SES). Measures used included household income, education, occupation, employment, poverty indicators, family composition, public assistance use, and having a computer in the household. In these studies, SES was used as a control variable. The proportion of each gender was equal (50% female).
Thirteen studies reported on the percentages of family meals that the study families shared.1,8–11,13,16–21 The majority of families had meals together 5 to 7 nights per week (52%), 31% shared 1 to 4 meals together, and 14% did not share any meals together.
Analyses were performed separately for cross-sectional and longitudinal studies. Some studies reported ORs for findings at baseline and at follow-up. In these cases, they contributed an effect size for each analysis. Fifteen studies reported cross-sectional findings,1,2,5,8–11,13–20 and 5 reported longitudinal findings.1,8,10,12,21
Eight studies1,2,8–13 examined family meals in relation to weight status. In terms of shared family meal frequency, studies considering overweight examined at least 3 meals per week compared with fewer than 3 meals per week. Four of the 8 studies reported nonsignificant findings.1,2,10,13 The studies included 44 016 participants, and individual study samples ranged from 145 to 14 431. Ages ranged from 4 to 17 years. The pooled OR was significant (0.88 [95% CI: 0.81–0.97]) with a test of heterogeneity (Q = 13.55; degrees of freedom [df] = 7; P = .06; I2 = 48.35%), which suggests that children and adolescents were 12% less likely to be overweight in families that had at least 3 shared family meals per week than those who ate fewer than 3 shared family meals per week. The fail-safe N revealed that there would have to be 50 studies located for the results to be nullified. The Q statistic and I2 value together suggest that the effect sizes do not come from the same population. Therefore, we tested age as a potential moderator, because the result of the test of heterogeneity was significant. There were 2 studies that examined younger children,9,12 2 that examined middle school–aged children,8,13 and 4 that examined older children (high school–aged).1,2,10,11 Age was not a significant moderator (QB = 1.32; df = 2; P = not significant).
There were 8 studies that examined food consumption.2,5,13–18 The consumption of unhealthy foods was analyzed and reported separately from the consumption of healthy foods for simplicity of interpretation of the ORs. The studies included a total of 56 919 participants (range: from 404–18 177 per study). Mean ages ranged from 2.8 to 17 years.
Six studies measured the consumption of unhealthy foods by adolescents.2,5,15–18 In terms of shared family meal frequency, studies compared at least 3 meals per week to fewer than 3 meals per week. The pooled OR was significant (0.80 [95% CI: 0.68–0.95]) with a test of heterogeneity (Q = 26.92; df = 5; P < .001; I2 = 81.43%), which suggests that children and adolescents in families that share at least 3 family meals per week have a 20% reduction in the odds of eating unhealthy foods than those in families that have fewer than 3 shared family meals together. The fail-safe N revealed that there would have to be 74 studies located for the results to be nullified. Because there was a significant Q value for heterogeneity, we tested age as a potential moderator. There were 2 studies that examined younger children (primary school– and junior high–aged)5,18 and 4 that examined older children (high school–aged).2,15–17 The difference between groups was significant (QB = 6.01; df = 1; P = .001); older children had a larger effect (OR: 0.74 [95% CI: 0.70–0.79]) than the younger children (OR: 0.93 [95% CI: 0.77–1.10]).
Five studies measured healthy dietary habits as well as the consumption of healthy foods,5,13–16 which included consumption of fruits and vegetables, eating breakfast, and taking a multivitamin. Studies that examined healthy foods and eating habits compared at least 3 to fewer than 3 shared family meals per week. The pooled OR was significant (1.24 [95% CI: 1.13–1.37]) with a test of heterogeneity (Q = 14.04; df = 4; P < .01; I2 = 71.51%), which suggests that families that shared at least 3 meals per week had children and adolescents who had an increased odds of 24% of eating healthy foods and maintaining healthy dietary habits than those whose families shared few or no family meals together. The fail-safe N revealed that there would have to be 101 studies located for these results to be nullified. Similar to unhealthy eating, the result of the test of heterogeneity was significant, so we tested age as a potential moderator. There were 2 studies that examined younger children (one which included infants14)5 and 3 that examined older children (middle school– and high school–aged).13,15,16 The difference between groups was marginally significant (QB = 6.75; df = 1; P = .06); older children had a larger effect (OR: 1.26 [95% CI: 1.20–1.33]) than the younger children (OR: 1.14 [95% CI: 1.07–1.21]).
Three studies examined shared family meals in relation to disordered eating.11,19,20 The studies included 104 353 adolescents (range: 145–99 462 per study). Mean ages ranged from 12.8 to 17.2 years. Each of the studies compared 5 or more versus ≤1 shared family meals per week. The pooled OR was significant (0.65 [95% CI: 0.58–0.73]) with a test of heterogeneity (Q = .98; df = 2; P = not significant; I2 = 0), which suggests that adolescents from families that share at least 5 meals per week are 35% less likely to engage in disordered eating than those that do not. The fail-safe N revealed that there would have to be 28 studies located for the results to be nullified.
There were 8 studies that measured shared family mealtime frequency as 5 or more meals versus ≤1 meal per week.8–11,16,18–20 We choose to run these analyses to contrast the extremes of eating the majority of meals versus virtually no meals together as a family. Five studies examined overweight, 2 examined disordered eating, and 2 examined unhealthy eating (1 study11 examined both overweight and disordered eating and, therefore, contributed a single effect). Outcomes from the 8 studies were pooled for this analysis to produce an overall summary effect. The question we asked is, “Does having regular shared family meals (≥5 per week) versus not having them influence nutritional health?” The answer is yes; the pooled OR was 0.75 (95% CI: 0.65–0.87). The odds are 25% less for children and adolescents that have 5 or more shared family meals together per week. The fail-safe N revealed that there would have to be 103 studies located for the results to be nullified. The result of the test for heterogeneity was significant (Q = 19.34; df = 7; P < .01; I2 = 64.28%). Because the result of the test for heterogeneity was significant, we ran a moderator analysis to examine the 3 outcome categories as a possible moderator. One study11 that examined both overweight and disordered eating contributed a separate effect size for both categories. The difference among groups was not significant (QB = 1.90; df = 2; P = not significant) (see Fig 2 for the forest plot, which graphically displays effect sizes for this analysis).
Five studies reported longitudinal results.1,8,10,12,21 Four of these studies examined overweight, and 1 examined disordered eating. Because just 1 study examined disordered eating, analyses were conducted only on the studies that examined overweight. Two of the studies had 5-year follow-ups,1,21 whereas the others reported on 3-year12 and 2-year8 follow-ups. In general, longitudinal studies were, individually, suggestive of little association between shared family mealtimes and outcomes. Of the 4 studies that reported on longitudinal findings, only 1 reported significant findings.12 However, the OR for the meta-analysis is significant (0.93 [95% CI: 0.90–0.95]), which suggests that shared family meals are associated with 7% odds of reduction of overweight and disordered eating. The result of the test for heterogeneity was not significant (Q = 1.94; df = 3; P = not significant; I2 = 0). The fail-safe N revealed that there would have to be 6 studies located for the results to be nullified.
To our knowledge, this is the first study to examine pooled estimates of risk across studies that have examined the association between shared family meals and various nutritional health outcomes for children and adolescents. The authors of 1 study22 performed a systematic review on healthy eating, but to our knowledge, none have performed a meta-analysis. Overall, families that eat 5 or more meals together have children who are ~25% less likely to encounter nutritional health issues than children who eat ≤1 meal with their families. Shared family meals seem to operate as a protective factor for overweight, unhealthy eating, and disordered eating. Although cross-sectional studies have revealed a stronger reduction in unhealthy behaviors relative to longitudinal studies, the longitudinal studies included in this meta-analysis were few and focused mainly on overweight; more longitudinal studies need to be conducted to shed light on the potential long-term relationship between family meals and nutritional health.
Although the mechanisms for how shared family meals relate to positive outcomes have not yet been empirically revealed, researchers have suggested several possibilities. It is not surprising that eating family dinners together is inversely associated with eating ready-made dinners, which feature lower nutrient values.15 For children or adolescents with disordered eating, mealtimes may provide a setting in which parents can recognize early signs and take steps to prevent detrimental patterns from turning into full-blown eating disorders. Indeed, dieting has been recognized as a precursor for the development of eating disorders.23 In addition, family meals are predictive of family-connectedness,24 which may encourage adolescents to talk about such issues within their families.
Adolescents themselves associate shared family meals with healthier eating.25 In a focus-group study, adolescents believed that they would eat healthier if they ate more meals with their families.25 This result is consistent with the age effects found in this meta-analysis, which demonstrate that adolescents who eat with their families eat healthier foods. In a study that asked adolescents about their interest in having their parents participate with them in a health-promotion program, adolescents stated that they would most like their parents to prepare healthy meals at home.26 It seems that there is interest by adolescents as well as receptivity in participating in family mealtimes, eating healthy foods, and learning about nutrition.
Four of the 8 studies that examined overweight children and adolescents reported nonsignificant findings. All 4 of these studies examined adolescents. Fulkerson et al1 suggested that shared family mealtimes may be more nutritionally beneficial for younger children because they are normative events, whereas for adolescents, shared family mealtimes decrease. We tested age as a potential moderator, but it was not significant, which may be because of the large variability we had in age. Two of our studies examined preschool/kindergarten-aged children, whereas the others overlapped middle school– and high school–aged children and adolescents. Similarly, 4 of the 5 longitudinal studies examined adolescents, whereas only 1 examined younger children. The study that examined younger children seemed to be pulling the longitudinal studies in a significant direction. The longitudinal finding was quite small and not consistent with the findings of the studies overall. More longitudinal studies are needed, specifically with younger children, to determine if the risk persists over time.
Our discussion assumes that family time is quality time. For some families, shared time is another opportunity for conflict. Indeed, observational studies of mealtime suggest that effective affect management, interpersonal involvement, and communication in families relate to child health.27 In a focus-group study, some adolescents reported not having shared family meals together because of dissatisfaction with family relations.25 Other barriers to sharing family meals together include parent work schedules, difficulty in planning ahead, and dealing with picky eaters.28,29 Health professionals are advised to tailor their recommendations on the basis of their knowledge about particular family situations. The families that do sit down to meals together may be more likely to be families that get along and thus contribute to the positive relations presented in this study, in both the short-term and long-term. However, there is emerging evidence to suggest that mealtime interactions that include positive forms of communication are not solely markers of overall family functioning but may represent a unique contribution to children's health and well-being.30
There is a large amount of variability in the studies conducted on family mealtimes. In particular, 5 studies did not report any information on SES. Investigating ethnic and gender differences was determined to be beyond the scope of this study. However, study results have indicated that this is an area that warrants further attention.10,19,21 The way in which nutritional outcomes were measured in the studies also varied. For example, some studies that examined obesity would classify a child with a BMI at the >85th percentile as being obese, whereas others would classify a child with a BMI at the >95th percentile as obese (eg,9,11). Furthermore, the question of how to measure family mealtimes has also varied from study to study. Some studies examined the extremes of shared family meal frequency, whereas others compared smaller increments, such as having 3 shared family meals per week versus less than 3. Several studies specifically asked about the frequency of shared family dinners, whereas others asked about shared family breakfasts and dinners, and others just asked more generally about shared family “meals.” In addition, even the definition of what constitutes “family” and who is present at the meal are frequently overlooked. Future research should include more precision in the measurement of not only the frequency of family mealtimes but structural aspects of the family and who is present during meals.
The results of this study suggest that shared family mealtimes offer nutritional benefits to family members. Health professionals are advised to encourage families to eat meals together. Future studies should further develop interventions for families that struggle with health issues such as obesity and disordered eating and add focus on family mealtimes as a setting in which to promote better nutrition habits. In addition, longitudinal studies can reveal any long-term potential that family meals may have in influencing nutritional health. Specific mechanisms of how family mealtimes influence related nutritional outcomes should be investigated.
This research was supported, in part, by the US Department of Agriculture National Institute of Food and Agriculture (Hatch Project number 793 328, to Dr Fiese).
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
*When the numbers of participants are reported throughout “Results,” they represent the true number of participants, which means that studies that shared the same sample are included only once.