Family Resemblance in Energy and Macronutrient Intakes: The Stanislas Family Study

Transcription

1 International Journal of Epidemiology International Epidemiological Association 1996 Vol. 25, No. 5 Printed in Great Britain Family Resemblance in Energy and Macronutrient Intakes: The Stanislas Family Study JEAN-MICHEL VAUTWER,* ANNE LLUCH," EDITH LECOMTE,* YVES ARTUR* AND BERNARD HERBETH* Vauthier J-M (Centre de Medecine Preventive, Vandoeuvre-les-Nancy, France), Lluch A, Lecomte E, Artur Y and Herbeth B. Family resemblance in energy and macronutrient intakes: The Stanislas Family Study. International Journal of Epidemiology 1996; 25: Background. There seems to be a consensus that family influences on dietary habits are important but few studies have addressed this issue directly. The purpose of this study was to determine if and how dietary intake aggregates within families. Methods. We examined the family aggregation of energy intake and the proportion of protein, fat and carbohydrate in the diet, estimated by a 3-day food consumption diary in 387 middle-class French families. Results. For energy and all macronutrients, spouse-spouse and child-child correlations were higher than parent-child correlations suggesting the minor contribution of genetics and the preponderant role of cultural and residual random environment. Variance component analysis confirmed the absence of genetic component for energy and all macronutrients and underlined the important role of a cohabitational effect for parents. Cultural inheritance represented 30-40% of dietary intake variance for children. Families who shared meals together more often had a lower residual random component. With the increasing number of meals eaten together (>45/week versus «45/week), between-generation components increased by about 10% for fat and carbohydrate, while for protein intake, the between-generation component for both parents (about 27%) and children (about 37%) remained unchanged. Conclusions. The general finding that dietary intake aggregates within families and that the individual behaviours are greatly influenced by characteristics within the family unit such as the number of meals eaten together provides additional justification for health promotion programmes that target the family as the unit for intervention. Keywords: family resemblance, energy intake, macronutrient intake, cultural inheritance It is well documented that risk factors for many diseases including coronary heart diseases, obesity, diabetes and some types of cancer occur within families. 1 " 3 While suggestive of a genetic effect, families also share cultural, behavioural and occupational factors that may also be related to disease risk. Dietary intake is one environmental factor implicated in the pathogenesis of chronic diseases and there seems to be a consensus that family influences on dietary habits are important. 4 " 9 Family resemblance in nutrient intake has been reported in spouses and between parents and their children; 10 " 15 however research on the similarity of dietary habits among family members is scarce and the extent to Centre de Medecine Preventive, Vandoeuvre-les-Nancy, France. *«Unite 308 INSERM, Nancy, France. Reprint requests to: Bernard Herbeth, Centre de Medecine Preventive, BP 7, 2 Avenue du Doyen Jacques Parisot, Vandoeuvre-les- Nancy Cedex, France. which this resemblance is genetically determined has not been clearly established. Studies on the food preferences of twins suggested a possible genetic component, children sharing genes with their biological siblings and parents, 16 " 21 but family studies did not agree with these results and underlined the preponderant influence of social environment. 22 " 25 During the first year of the Stanislas Family Study, we collected detailed dietary data on parents and their children aged 7-25 years. As a preliminary step, we investigated the family aggregation of energy intake and of the proportion of macronutrients in the diet among these French nuclear families. First, we computed the family correlations for energy and macronutrients. Secondly, variance component analysis was applied in order to assess the relative contributions of genetics, common household factors and the individual specific environment. Thirdly, we detailed the variance component analysis according to the number of meals shared by relatives. 1030

2 FAMILY RESEMBLANCE IN ENERGY AND NUTRIENT INTAKE 1031 METHODS Subjects The current analysis uses data from the Stanislas Family Study, a longitudinal survey involving 1000 families and designed to investigate the family factors related to cardiovascular diseases. A subgroup of 387 nuclear Caucasian families from the Nancy area was recruited at the Centre for Preventive Medicine of Vandoeuvre-les-Nancy (France) between January 1994 and May 1995 to have a free health check-up. Only volunteer families composed of two parents aged less than 65 years and two biological children older than 7 years were included. Families were excluded from participation if one member had a disease that could influence eating behaviour. The protocol and the aim of the study were fully explained to the subjects who gave their written consent. The research protocol was approved by the Scientific Ethics Committee of Nancy. Dietary Record To estimate nutrient intake, the 774 parents and 774 children completed a 3-day food consumption diary on two contiguous week days and one weekend day assigned at random. and children were assigned the same 3 days. Participants were given formal instruction on keeping accurate records without changing their normal eating habits, and the completed diaries were carefully reviewed by the dietitian using a portion size picture booklet. 26 Each subject was responsible for his own recording; for young children only diaries were completed by the mother, with help from the child. The 3-day records were entered and analysed by using a commercially computerized, nutrition data base program. 27 The nutrient values in the database include the most recent available French data and food-industry values. 28 The dietary variables analysed in this report were the means of the different variables reported over the 3-day period. During the family interview, information was collected about schedules of meals, location of eating, social environment of meals, and for breakfast, lunch and dinner, the number of meals per week shared by each pair of relatives (mother-father, mother-child 1, mother-child2, father-child I, father-child2, child 1- child2); the sum of the number of meals shared by these six pairs of relatives, for the three principal meals, was used as an index of similarities. To examine the effect of this index on familial aggregation, our sample of families was dichotomized at the median of the distribution (45 meals per week eaten together). Statistical Analysis The means and standard deviations for daily intakes of energy and macronutrients were calculated for mothers, fathers, sons and daughters and the distributions of all variables were tested for normality. Two classical methods were used to determine family resemblance of dietary intakes: correlation coefficient computation and variance component analysis. Since energy and macronutrient intakes were significantly correlated with age, the two methods were used on age-adjusted dietary data. The two statistical programs allowed adjustment for age within models, simultaneously and separately for fathers, mothers and offspring. The family correlations were jointly computed using the maximum likelihood method described by Donner and Koval. 29 Differences in family correlations according to parent's or child's sex were tested by comparing the log-likelihood of different nested models. If LI is the log-likelihood of a general model and L2 the loglikelihood of a submodel, 2 (L1-L2) has an approximate X 2 distribution with degrees of freedom equal to the number of parameters restricted when specifying the submodel from the general model. 29 ' 30 Then, a multivariate normal model for pedigree analysis was considered using the approach of Lange et a/. 31 ' 32 The general model assumes that the studied trait (dietary intake) is the result of the sum of three independent random components: a polygenic component 'G' (representing additive genetic factors), a common environmental factor 'C (factors shared by members of the same family) and a residual 'E' (representing environmental factors specific to each individual). Each component is assumed to be random with mean zero and variance <J 2 G, o 2 c, and o 2 E, respectively. The total variance of the trait is decomposed as a 2 = a 2 G + o 2 c + o 2 E. The estimates of these parameters are based on calculation of covariance matrices accounting for kinship coefficients and result from application of the method of maximum likelihood, using the computer package FISHER 31. This program also performs tests of fit of the underlying multinormal distribution. Comparison of nested models was based on the likelihood ratio criteria. After having observed the absence of a significant polygenic component, a more complex model was tested: o 2 c (common environmental component) was subdivided into three components resulting from a common spouse environment (o 2 sp), a parent-child environment (c 2^) and a child-child common environment (o 2 cc) taking into account the number of meals shared by the family members. Each component was estimated separately according to the cumulative number of the three principal meals shared per week by all pairs of relatives (^45 or 3=45 meals shared per week). For spouses and offspring, respectively, this model allowed estimation of three components of

3 1032 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY TABLE 1 Age, Quetelet index, daily energy and macronutrient intakes of parents and offspring" Fathers (n = 387) Mothers (n = 387) Sons (n = 365) Daughters (n = 409) Age (years) Quetelet index (kg/m 2 ) Nonalcohol energy (kcal) Carbohydrate (g) Fat(g) Protein (g) Carbohydrate (%) b Fat (%)" Protein (%) b 42.4 ± ± ± ±77 99 ± ± ± ± ± ± ± ± ± ±21 78 ± ± ± ± ± ± ± ± ±28 94 ± ± ± ± ± ± ± ± ±22 73 ± ± ± ±2.8 a x±sd. b Percentage of nonalcohol energy. TABLE 2 Family correlations for energy and macronutrient intakes adjusted for age Energy (kcal) Protein (%f Fat (%)' Carbohydrate (%) a GM b SM C GM b SM C GM b SM C GM b SM C FM (n = 387) FS (n = 365) FD (n = 409) MS (n = 365) MD (n = 409) SS (n = 93) SD(n= 179) DD(n= 115) - Log, L d f S NS NS NS * Percentage of non-alcohol energy. b GM, general model estimating all the correlations; FM, father-mother; FS, father-son; FD, father-daughter; MS, mother-son; MD, mother-daughter; SS, son-son; SD, son-daughter; DD, daughter-daughter. c SM, submodel estimating only three correlations without accounting for parent's and child's sex; FM, father-mother; FS = FD = MS = MD, parentchild; SS = SD = DD, child-child. d Logarithm of likelihood function. c Tests for parent's and child's sex differences X 2 (5 d.f.). variance: between-generations or cultural inheritance 2 within generations (a 2 sp or o 2 cc) and residual p spp random variance (o 2 E) the sum of these two last components being the non-transmissible variance. RESULTS Mean characteristics and daily nutrient intakes of parents and offspring are presented in Table 1. While age and Quetelet Index were lower for mothers than for fathers (P < 0.001), no difference was found between sons and daughters. Mean energy, carbohydrate, fat, and protein intakes were higher in males than in females for both parents and offspring. However, when expressed as per cent of non-alcohol energy intake, differences between males and females disappeared. The pattern of family correlations for energy intake and the proportion of macronutrients in the diet are given in Table 2. Significant correlations (P < 0.001) were found for all the various pairs of relatives. For energy and the three macronutrients, spouse-offspring correlation coefficients were lower than those between

4 FAMILY RESEMBLANCE IN ENERGY AND NUTRIENT INTAKE 1033 the other pairs of relatives. Father-offspring correlations were higher than mother-offspring correlations for all parameters but when testing for sib's or parent's sex difference, the test only reached statistical significance for energy intake. Then, the contribution of genetic and environmental factors was investigated using the variance component model. No significant genetic contribution to variability in energy and macronutrient intakes was found. A mode! including only shared and individual-specific environmental effects, without genetic contribution, was considered to give the best description of the composition of energy and macronutrient variance. Shared common environment accounted for 32, 37, 37 and 33% of the variance in energy, fat, protein and carbohydrate, respectively (data not shown). As a consequence of the previous results, a more complex model, presented in Table 3, was tested. Shared common environment (c 2 c) was subdivided into three components as described in Methods. For energy intake, model 3 did not differ significantly from model 1. As the likelihood function was higher for model 3, it better described the composition of variance in energy intake. For macronutrient intake, significant differences were found between the full model and other models. So, the full model including shared and individual-specific environmental effects and number of meals eaten at home was considered to give the best description of variation in macronutrient intake. For energy and macronutrient intake, total phenotypic variance was lower when meals eaten together were more frequent. Components of variance (in per cent) for spouses and offspring are presented in Table 4. For energy, shared common environment (between and within-generations) represented about 32% of the total phenotypic variance whatever the number of meals was. For all nutrients, the residual random component (o 2 E) decreased for parents and offspring when numbers of meals eaten at home increased. For parents, the within-generation component (o" 2 sp) was the major component, representing about 45% of total phenotypic variance; o 2 sp increased when the number of meals eaten together was S=45 meals/week. For children, both within (O 2^) and between (O 2 CC) generation components were important and both increased by about 10% with increasing number of meals eaten at home. For parents and children, between-generation components increased by about 15% and 12% for fat and carbohydrate respectively, with increasing number of meals eaten together. However for protein intake the number of meals shared within the family did not influence the between-generation component for both parents (about 27%) and children (about 37%). DISCUSSION This study is the first to examine the family aggregation of energy and macronutrient intakes in a cohort of middle-class French families, taking into account the number of meals shared. No significant genetic effect was found for intake of energy and nutrients tested, and cultural inheritance was found to represent 30-40% of the total phenotypic variance, depending on the number of meals shared. Like Perusse et al. 12 we used the 3-day dietary record to estimate nutrient intake in our sample of relatives and we also questioned the validity of this methodology. However previous studies showed that the 3-day dietary record including a weekend day provided reliable estimates of nutrient intake that were comparable to those derived from a 7-day record. 33 " 35 Correlation coefficients between relatives were relatively modest for all nutrients in the reduced model using the maximum likelihood method described by Donner. 29 For energy and all macronutrient intakes we found that between-generation correlations were lower than within-generation correlations suggesting that the contribution of specific common factors for parents and children, respectively, was preponderant. These low parent-child correlations were previously mentioned by Rozin et al. and called 'the family paradox'. 36 The moderate levels of family aggregation of dietary variables found in this study are generally in agreement with previous reports. l0 ~ 15 Oliveria et a/. 10 showed that most of the nutrients were more closely correlated between parents than between parents and children, confirming data from Perusse et al} 2 Spouse correlations ranged from 0.25 to 0.50 while parent-child correlations ranged from 0.25 to l012 Analysis using the FISHER program showed that best models for our family data were models with no genetic influence. For all nutrients, common environmental effects represented about 32% of the total phenotypic variation. For macronutrients, after eliminating the genetic component, parent-child environmental component represented 28-46% of children's total phenotypic variance. Although Perusse et al} 2 found a little biological inheritance for macronutrient intake ranging from 11% to 20%, this genetic effect was not acceptable in the absence of a statistically valid solution. In agreement with our results they found a transmissible variance ranging from 31% to 45% and concluded that nonbiological factors, and particularly environmental factors non-transmissible between generations, were the preponderant factors of the phenotypic variation. Most twin studies support our results by suggesting that food preferences are predominantly environmental,

5 1034 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY TABLE 3 Variance components of energy and macronutrient intakes according to number of meals eaten at home by parents and children adjusted for age" Model 1 =S45 meals/week Model 2 _ 2 b spa spb p 2 EA = Model _2 b spa pca cca ' spb ~ ^ pcb ~ ^ ccb i meals/week Energy (kcal) Comparison _ vs 1 S vs 1 NS Protein (%) E - Log. L< Comparison > _ vs 1 s vs 1 =s Fat (%) \ - Log, L< Comparison /*> vs 1 s vs 1 S Carbohydrate (%) - Log e L c Comparison _ vs 1 = vs 1 s a <* 2 sp. spouses common environment component; o" 2, parent-child common environmental component; o* 2 cc, child-child common environmental component; O 2 E, residual random environment component. b A, lower than 45 meals/week; B, greater than 45 meals/week. 0 Logarithm of likelihood function. d Likelihood ratio test. and not genetically determined. 16l8 ' 20>21 The exception to this was De Castro, 37 who found that overall nutrient intake is a highly heritable characteristic and, given that overall intake is a product of the size and frequency of meals, there may also be a high degree of heritability in characteristics of the meal pattern. However, De Castro used the Holzinger Index of Heritability, the validity of which as a coefficient of heritability was questioned by Christian el al? % who demonstrated that some traits may show a major difference in total variance between dizygotic and monozygotic twins. Falsely high estimates of genetic influence may have resulted from unequal

6 FAMILY RESEMBLANCE IN ENERGY AND NUTRIENT INTAKE 1035 TABLE 4 Variance components expressed in per cent for spouses and offspring Between-generations component Within-generation component (a 2 sp for parents and a 2 for offspring) Residual component Energy (kcal) =s45 meals/week (32.0)' (32.0)" (32.3)' (32.3/ Protein (%) S45 meals/week Fat (%) = 45 meals/week Carbohydrate (%) = 45 meals/week a For energy, within and between-generation components were estimated together. environmental effects between dizygotic and monozygotic twins. 18 The extent to which genetic factors influence food consumption evidently requires further investigation. Taking into account the number of meals eaten together within the family enabled more precise definition of the different components of shared environment. First, the preponderant part of parent-parent common environment component of variance and secondly the equal part of parent-child and child-child components of variance tended to confirm the importance of the within-generation component and the lack of importance of a common environment between generations. So, the role of cultural inheritance was minor, although families who share meals more often have a greater shared parent-child environment component. Two studies 39 ' 40 manipulated the number of relatives present at a meal and suggested that the effect of the presence of relatives by extending meal duration on energy intake is a causal effect. We also found that spouses exerted a moderate but definite mutual influence on dietary intakes. These spouse correlations were attributable to a mutual influence from communal living or to a tendency of like to mate with like (i.e. assortative mating), or both; this hypothesis remains to be confirmed In the literature, many factors have been identified which contribute to dietary intake similarities within

7 1036 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY the family (e.g. mutual influence, food preferences, eating behaviours, family organization, atmosphere and duration of meals) and particularly between spouses and children (e.g. adult role modelling, parent discipline, parent-child stimulation). 25 ' As a first step, we used a global index of family environment, the number of shared meals by relatives. Using our longitudinal sample of French families, future investigations will be conducted taking into account more accurate environment and cultural indices, in order to characterize the determinants of food and nutrient resemblance better. Nevertheless, the general finding that dietary intake aggregates within families and that individual behaviour is greatly influenced by characteristics within the family unit such as the number of meals eaten together, provides additional justification for health promotion programmes that target the family as the unit of intervention. 43 ACKNOWLEDGEMENTS We thank the management, reception and preclinic staff of the Center for Preventive Medicine of Vandoeuvreles-Nancy (France) for their help during this study. We are especially grateful to Sylvie Pechine for the collection of food intake data, Mayrvonne Chaussard and Chantal Lafaurie for the family recruitment and Ren6 Gue'guen for his help and advice in the statistical analysis. We are also indebted to the families of the Stanislas survey who made this study possible. We would like to thank also Laurence Tiret and Roger Rakotovao (INSERM U258, Paris, France) for providing the program package of familial correlation computation. REFERENCES 1 Sellers T A, Kushi L H, Potter J D. Can dietary intake patterns account for the familial aggregation of disease? Evidence from adult siblings living apart. Genet Epidemiol 1991; 8: Perkins K A. Family history of coronary heart disease: Is it an independent risk factor? Am J Epidemiol 1986; 124: Austin M A, King M-C, Bawol R D, Hulley S B, Friedman G D. Risk factors for coronary heart disease in adult female twins: Genetic heritabilily and shared environmental influences. Am J Epidemiol 1987; 125: Lissner L, Heitmann B L. Dietary fat and obesity: evidence from epidemiology. Eur J Clin Nutr 1995; 49: Sallis J F, Broyles S L, Franck-Spohrer G, Berry C C, Davis T B, Nader P R. Child's home environment in relation to the mother's adiposity. Int J Obesity 1995; 19: Potter J D, Graves K. Diet and cancer: Evidence and mechanisms an adaptation argument. In: Rowland I (ed.). Nutrition and Toxicology. New Jersey: The Telford Press, Carroll K K. Dietary fats and cancer. Am J Clin Nutr 1991; S3: 1064S-67S. 8 Motulsky A G. Nutrition and genetic susceptibility to common diseases. Am J Clin Nutr 1992; 55: 1244S-45S. 9 Connor S L, Connor W E, Henry H, Sexton G, Keenan E J. The effect of familial relationships, age, body weight, and diet on blood pressure and the 24 h urinary excretion of sodium, potassium, and creatinine in men, women, and children of randomly selected families. Circulation 1984; 70: Oliveria S A, Ellison R C, Moore L L, Gillman M W, Garrahie E J, Singer M R. Parent-child relationships in nutrient intake : the Framingham Children's Study. Am J Clin Nutr 1992; 56: " Patterson T L, Rupp J W, Sallis J F, Atkins C J, Nader P R. Aggregation of dietary calories, fats, and sodium in mexican-american and anglo families. Am J Prev Med 1988; 4: Perusse L, Tremblay A, Leblanc C et at. Familial resemblance in energy intake: contribution of genetic and environmental factors. Am J Clin Nutr 1988; 47: Laskarzewski P, Morrison J A, Khoury P et at. Parent-child nutrient intake interrelationships in school children aged 6 to 19: The Princeton School District Study. Am J Clin Nutr 1980; 33: Kolonel L N, Lee J. Husband-wife correspondance in smoking, drinking, and dietary habits. Am J Clin Nutr 1981; 34: Kolonel L N, Lee J. Nutrient intakes of husbands and wives: implications for epidemiologic research. Am J Epidemiol 1981; 115: Rozin P, Millman L. Family environment, not heredity, accounts for family resemblances in food preferences and attitudes: a twin study. Appetite 1987; 8: Falciglia G A, Norton A. Evidence for a genetic influence on preference for some foods. J Am Diet Assoc 1994; 94: Fabsitz R R, Garrison R J, Feinleib M, Hjortland M. A twin analysis of dietary intake: evidence for a need to control for possible environmental differences in MZ and DZ twins. Behav Genet 1978; 8: Heller R F, O'Connell D L, Roberts DCKel al. Lifestyle factors in monozygotic and dizygotic twins. Genet Epidemiol 1988; 5: Faust J. A twin study of personal preferences. J Biosoc Sci 1974; 6: Greene L S, Desor J A, Mailer O. Heredity and experience: their relative importance in the development of taste preference in man. J Comp Physiol Psychol 1975; 89: Birch L L. The relationship between children's food preferences and those of their parents. J Nutr Educ 1980; 12: Borah-Giddens J, Falciglia G A. A meta-analysis of the relationship in food preferences between parents and children. J Nutr Educ 1993; 25: Logue A W, Logue C M, Uzzo R G, McCarty M J, Smith M E. Food preferences in families. Appetite 1988; 10: Pliner P, Pelchat M L. Similarities in food preferences between children and their siblings and parents. Appetite 1986; 7: Deheeger M, Le Moullec N, Monteiro P. Preziosi P, Hercberg S. Validation du document iconographique realise pour l'enquete alimentaire de I'etude Suvimax. Inform diit 1994; 2:

8 FAMILY RESEMBLANCE IN ENERGY AND NUTRIENT INTAKE Musse N, Michaud C, Musse J P, Nicolas J P. Gestion informatisee de l'enquete alimentaire. Montreal, XII'"" Congres International de Medecine Sociale, 1989: p. 53 (abstr). 28 Feinberg M, Favier J C, Ireland-Ripert J. Repertoire giniral des aliments. Table de composition. FFN CIQUAL. In: INRA Lavoisier, Tech et Doc, Paris, Donner A, Kovall J. A multivariate analysis of family data. Am J Epidemiol 1981; 114: Rao D C, Vogler G P, McGue M, Russel J M. Maximumlikehood estimation of familial correlations from multivariate quantitative data on pedigrees: a general method and examples. Am J Hum Genet 1987; 41: Lange K, Weeks D, Boehnke M. Programs for Pedigree Analysis: MENDEL, FISHER, and DGENE. Genet Epidemiol 1988; 5: Lange K, Westlake J, Spence M A. Extensions to pedigree analysis III. Variance components by the scoring method. Ann J Hum Genet 1976; 39: Tremblay A, Sevigny J, Leblanc C, Bouchard C. The reproducibility of a three-day dietary record. Nutr Res 1983; 3: Bingham S A, Nelson M, Paul A A, Haraldsdottir J, Loken E B, Van Staveren W A. Methods for data collection at an individual level. In: Cameron M E and Van Staveren WA (eds). Manual on Methodology for Food Consumption Studies. Oxford: Oxford University Press, 1988, pp Stuff J E, Garza C, Smith E O, Nichols B L, Montandon M. A comparison of dietary methods in nutritional studies. Am J Clin Nutr 1983; 37: Rozin P. Family resemblance in food and other domains: The family paradox and the role of parental congruence. Appetite 1991; 16: De Castro J M. Genetic influences on daily intake and meal patterns of humans. Physiol Behav 1993; S3: Christian J C, Kang K W, Norton J A. Choice of an estimate of genetic variance from twin data. Am J Hum Genet 1974; 26: Feunekes G I J, Graaf C, van Staveren W A. Social facilitation of food intake is mediated by meal duration. Physiol Behav 1996 (In press). 40 Redd M, De Castro J M. Social facilitation of eating, effects of social instruction on food intake. Physiol Behav 1992; 52: Hertzler A A. Children's food patterns a review. II Family and group behavior. J Am Diet Assoc 1983; 83: Schaffer R B, Keith P M. Influences on food decisions across the family life cycle. J Am Diet Assoc 1981; 78: Baranowski T, Nader P R. Family involment in health behavior change programs. In: Turk D C, Kerns R D (eds). Health, Illness, and Families. New York: John Wiley, 1985, pp (Revised version received March 1996)