AJCN Tufts Nutrition Symposium, Boston Sept 24-26
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Purchase Article
Right arrow View Shopping Cart
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by van den Bree, M. B.
Right arrow Articles by Dwyer, J. T
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by van den Bree, M. B.
Right arrow Articles by Dwyer, J. T
Agricola
Right arrow Articles by van den Bree, M. B.
Right arrow Articles by Dwyer, J. T
American Journal of Clinical Nutrition, Vol. 70, No. 4, 456-465, October 1999
© 1999 American Society for Clinical Nutrition

Original Research Communications

Genetic and environmental influences on eating patterns of twins aged >=50 y1,2,3,4

Marianne BM van den Bree, Lindon J Eaves and Johanna T Dwyer

1 From the Virginia Institute for Psychiatric and Behavioral Genetics, Virginia Commonwealth University, Richmond, and the Tufts University Schools of Medicine and Nutrition, Jean Mayer Human Nutrition Research Center on Aging at Tufts University and Frances Stern Nutrition Center, New England Medical Center, Boston.

2 The contents of this publication do not necessarily reflect the views or policies of the US Department of Agriculture, nor does mention of trade names, commercial products, or organizations imply endorsement by the US government.

3 Supported by grant AA-06781 from the National Institutes of Health (National Institute of Alcohol Abuse and Alcoholism) and gifts from RJR Nabisco corporations and the Gerber Foundation and based on work supported by the US Department of Agriculture under agreement 58-1950-9-001.

4 Address reprint requests to MBM van den Bree, National Institute on Drug Abuse, Intramural Research Program, National Institutes of Health, PO Box 5180, Baltimore, MD 21224. E-mail: mvandenb{at}intra.nida.nih.gov.


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 SUBJECTS AND METHODS
 RESULTS
 DISCUSSION
 APPENDIX A
 REFERENCES
 
Background: Clinicians and researchers could benefit from a greater understanding of the role of genetic and environmental factors in human eating behavior.

Objective: Our aim was to estimate the relative influence of genetic and environmental factors on habitual eating patterns in middle-aged and elderly men and women.

Design: Male and female twins (n = 4640) aged >=50 y completed a mailed version of the National Cancer Institute food-frequency questionnaire. Factor analysis was performed to identify eating patterns among respondents. Estimates of genetic, common environmental (shared by family members), and specific environmental (unique to an individual) influences were obtained for food use, serving size, and consumption frequency by comparing monozygotic and dizygotic twin-pair groups with structural equation analysis.

Results: Two independent eating patterns were identified: the first consisted of items high in fat, salt, and sugar, and the second reflected healthful eating habits. Although the influence of environmental factors was larger, between 15% and 38% of the total variation in pattern 1 and between 33% and 40% in pattern 2 were explained by genetic influences. Models accounting for sex differences in genetic and environmental estimates fit the data significantly better for food use and serving size of foods in eating pattern 1 and for food use in eating pattern 2.

Conclusion: Although 60–85% of the variability in eating patterns was associated with environmental factors, genetic influences were also apparent and there was some evidence of sex specificity. These findings may be important in crafting dietary interventions and predicting adherence to these interventions.

Key Words: Twins • semiquantitative food-frequency questionnaire • eating patterns • genetic influences • environmental influences • sex • eating behavior


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
SUBJECTS AND METHODS
 RESULTS
 DISCUSSION
 APPENDIX A
 REFERENCES
 
Several studies have indicated that the human diet can be described in terms of a limited number of eating patterns (1–8). Most (1–7) but not all (8) studies reported healthful eating patterns, characterized by the consumption of fruit, vegetables, whole grains, dietary fiber, and milk and milk products, and less healthful eating patterns, characterized by the consumption of alcohol and foods high in fat, sugar and salt and that may predict later diet-related disease.

Environmental influences on food choice and eating patterns are well recognized, but genetic factors may also be involved. Family members show similarities in food preferences (9–14). However, genetically informative study designs are necessary to separate the influences of genetic factors and the family environment. Most early twin studies had limitations in sample size, the dietary measures used, or the method of analysis used. Nevertheless, most (15–21) but not all (22, 23) twin studies published to date indicate that eating behavior is influenced by genes. However, results are inconsistent with respect to the effect of genetic factors. The variation in estimates of genetic influences across studies may be attributable to the dietary measures used. Genetic influences may be stronger for nutrients than for individual food items (15) and for food selection measures rather than measures of absolute food intake (19). These findings suggest that patterns of intake (eg, the interrelations between dietary items) may be more sensitive to genetic influences than are relatively independent measures of consumption. Differences in sample characteristics (including age and sex) may also influence findings. For example, genetic influences are more difficult to detect in samples with wide age ranges when genetic factors influencing diet act predominantly during a certain portion of the life span (eg, adolescence or older age). Similarly, when genetic and environmental factors differ between the sexes, results may be influenced by the proportion of men to women in the sample.

Another issue that may influence results of twin studies of eating behavior is violation of the equal environment assumption. The twin method assumes that the effect of the environment on a trait is identical for monozygotic and dizygotic twins. If this assumption is violated, greater resemblance of monozygotic than dizygotic twins, which is usually attributed to genetic factors, could be due to environmental influences (24). Two twin studies indicated that monozygotic twins may have more similar food intakes because they have more frequent contact than dizygotic twins (15, 19), whereas a third study failed to replicate these findings (20). However, it is also possible that greater contact may not increase twin similarity, but rather that twins who are more similar seek each other's company more frequently (25). Because the issue is unresolved at this point, twin studies of food intake should test the possibility that heritability estimates may be inflated because the equal environment assumption is violated.

The first aim of the present study was to establish habitual eating patterns in a large sample of twins. The second aim was to estimate the relative influence of genetic and environmental factors on these patterns. A final aim was to establish whether significant sex differences existed in the genetic and environmental factors underlying the eating patterns.


   SUBJECTS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 SUBJECTS AND METHODS
RESULTS
 DISCUSSION
 APPENDIX A
 REFERENCES
 
Subject sample
Twins (n = 8289) residing throughout the United States were recruited through an advertisement in the Journal of the Amercan Association of Retired Persons and subsequently enrolled in a twin registry as part of a larger study examining the role of genes and the environment in attitudes, behaviors, and health. The study was approved by the Human Subjects Committee at Virginia Commonwealth University. Twins who had returned an earlier questionnaire were contacted again between August 1989 and March 1991 and invited to complete a questionnaire on diet and related characteristics. In response to this solicitation, 5579 (67%) of those contacted agreed to participate and returned the questionnaire described below. Of these, 939 (17%) were excluded because they were <50 y old. Therefore, food-frequency questionnaire data were available for 4640 persons (3378 women and 1262 men). Zygosity information was available for 1648 complete twin pairs (71% of 4640 subjects): 210 monozygotic male pairs, 725 monozygotic female pairs, 92 dizygotic male pairs, 376 dizygotic female pairs, and 245 dizygotic opposite-sex pairs. A group of 218 respondents (113 women and 105 men) completed the dietary questionnaire a second time, 0.5–1 y after completing the first questionnaire. These subjects formed the sample on which test-retest analyses were based.

Instruments
The National Cancer Institute's (26) semiquantitative food-frequency questionnaire (SFFQ) was used to obtain dietary information. The SFFQ was derived from a large, representative, noninstitutionalized sample of the US population surveyed in the second National Health and Nutrition Examination Survey of 1976–1980. The questionnaire includes foods and food groups contributing the greatest amounts of food energy and certain nutrients to the American diet. For each of 99 items on the SFFQ, use of food or food group (yes or no), consumption frequency (number of times per day, week, month, and year), and usual serving size (small, medium, or large) over the past year are queried. The reliability and concurrent validity of the SFFQ are well established (27–30). Information on the zygosity of twins was gathered with 2 questions about physical similarity (resemblance as children and being confused by others as children). These items have been shown to provide >90% accuracy in ascertaining zygosity (31, 32).

Analysis of eating patterns
Factor analysis was used to identify patterns in the relations among the foods and food groups. The number of factors retained was decided from scree plots of eigenvalues of principal components and by subjectively evaluating meaningful patterns (33). Agreement between factor patterns derived from subgroups of the sample was established by calculating coefficients of congruence (CC) and root-mean-square (RMS) deviations from principal components (34).

For each subject, individual eating pattern scores were calculated by adding those food use and nonuse items with relatively high loadings on a factor (>=0.30). Factors were first rotated (Varimax solution) to make them independent. Thus, for each subject, a score was obtained based on foods representing the first factor, and another independent score was obtained based on foods representing the second factor. The total number of items used in the calculations was 34 for eating pattern 1 and 36 for eating pattern 2. The same items were also used to calculate eating patterns based on serving size and consumption frequency information. When consumption frequency information was specified per day, week, or month, the data were converted into consumption of each food per year. Responses were subsequently converted to a 5-point, rank-order scale to reduce the wide variation in responses. The eating patterns were further characterized by calculating polyserial correlation coefficients (35) with other diet-related items appearing on the SFFQ (eg, frequency of eating the skin on chicken, eating the fat on meat, adding salt to food, and taking vitamin and mineral supplements—all categorical data).

Eating pattern scores were coded as missing if there were >=10 missing values for the 99 foods and food group items or if there were nonrandom missing values (eg, that might be caused by skipping a page). For subjects with <10 random missing values, eating pattern scores were imputed by replacing a missing item with the mean of the nonmissing responses. Means and SDs of the imputed scores were relatively small. For pattern 1 foods they were as follows: food use, 0.14 ± 0.46; serving size, 0.86 ± 1.16; and consumption frequency, 0.39 ± 0.78. The scores for pattern 2 foods were as follows: food use, 0.16 ± 0.48; serving size, 1.08 ± 1.31; and consumption frequency, 0.51 ± 0.91. When this procedure was followed, dietary information was available for the following numbers of respondents: food use or nonuse, 4283 (92% of original sample); serving size, 3191 (69%); and food consumption frequency, 3931 (85%). The significance of test-retest differences in mean eating pattern scores was evaluated with paired t tests as well as by Pearson correlation analysis. These analyses were performed for men and women separately.

Genetic theory
Similarities between biological relatives may be due to either genetic influences or common environmental influences. For twin pair members, common environmental influences could include the neighborhood the twins grew up in, their family's rearing practices, and friends that were shared. Specific environmental influences, on the other hand, include all influences not shared between twin pair members. These influences have differential effects on each twin and tend to make them less alike.

Monozygotic twins are genetically identical and, if they are reared together, have a common family environment. Dizygotic twins who are raised together also share the same family environment but have on average only half of their genes in common. If certain assumptions apply—ie, monozygotic and dizygotic twins are equally affected by their environments (equal environment assumption), mating is random, and gene-environment interactions or correlations are absent—then monozygotic twin correlations for eating habits that are about twice the size as those for dizygotic twins suggest that genetic influences are present and that they are additive in nature. A dizygotic correlation greater than half the monozygotic correlation suggests common environmental influences, whereas a monozygotic correlation <1.0 suggests that specific environmental influences are present.

Different correlation patterns for male and female monozygotic and dizygotic twin-pair groups suggest that sex differences exist in genetic influences, environmental influences, or both. Lower correlations in dizygotic opposite-sex than in dizygotic same-sex twins also suggest sex differences. The principles of twin methodology are outlined in detail in several sources (36, 37).

Genetic analyses
The effects of age and sex were removed from the eating pattern means by using regression analysis. Regression residuals were subsequently normalized by using normal rank scores (38). Pearson correlation coefficients were calculated for each measure individually for the twin-pair groups (monozygotic male pairs, monozygotic female pairs, dizygotic male pairs, dizygotic female pairs, and dizygotic opposite-sex pairs). Correlation coefficients were converted to standardized random variables to assess whether correlations of monozygotic twin pairs were significantly greater than those of dizygotic twin pairs (one-tailed test) (39). Subsequently, variance-covariance matrices were calculated for each measure for the 5 twin-pair groups and were the input for the genetic analyses. To determine whether there were sex differences in the genetic and environmental contributions to the eating patterns, dizygotic opposite-sex twin pairs were arranged so that women were always designated twin 1 and men twin 2.

Estimates of genetic and environmental contributions to the eating patterns were obtained by structural equation analysis with the program MX (40). Four different models were used to test hypotheses of genetic and environmental influences on eating patterns. The basic model included additive genetic influences (A), as well as common and specific environmental influences (C and E, respectively), and assumed that these were the same for both sexes (ACE model). Three other models were used to test specific hypotheses of sex differences in genetic and environmental influences (36, 41). The common-effects sex-limitation model assumed that the same genes and environmental factors influenced the eating patterns of both men and women, but that the magnitude of the influences differed for the sexes. Hence, this model estimated sex-specific contributions of additive genetic as well as common and specific environmental influences for men (Am, Cm, and Em) and women (Af, Cf, and Ef). Similar to the common-effects sex-limitation model, the general-effects sex-limitation model also estimated sex-specific parameters for men and women (Am, Cm, and Em, and Af, Cf, and Ef). In addition, however, this model also used information from opposite-sex twin pairs to test the possibility that eating patterns were influenced by different genetic factors (or environmental factors) for men and women. In the genetic variant of the model, the genetic correlation (rg) between opposite-sex twin pairs was estimated (range: 0–0.5) rather than being fixed at 0.5. In the environmental variant of the model, the common environmental correlation (rc) was estimated (range: 0–1.0) rather than being fixed at 1.0. Finally, the scalar sex-limitation model assumed that identical genes (and environmental factors) influenced eating patterns in both sexes, but that they had different effects on the total variation in eating behavior. The model included a scalar (k), which was estimated, and accounted for sex differences by multiplication with the total phenotypic variation for one sex, but not the other. Unlike the other models, this model was empirical rather than based on biological assumptions. In all models, in addition to specific environmental influences, E also included measurement error. The relative contributions of genetic influences (h2, or heritability estimate), as well as common (c2) and specific environmental influences (e2), were obtained by dividing their variation by the total phenotypic variation for the eating pattern measure (genetic and environmental influences combined).

Models were fitted by the maximum likelihood method. The fit of the model was evaluated by the P value of the associated chi-square statistic. A small P value suggested considerable discrepancy between the model and the observed data, whereas larger values indicated a better fit. Another indicator of fit is Akaike's information criterion (AIC) (calculated as chi-square – 2 x df) (42). When the models were compared, low AIC values indicated a more parsimonious explanation of the data. The significance of sex differences was tested by comparing the fit of each of the sex-specific models with that of the basic ACE model. Results are presented for the full model, including A, C, and E. The significance of the individual parameters can be evaluated by their 95% CIs, which were calculated by using MX (40).

Equal environment assumption
The association between the food pattern measures and indicators of childhood closeness, adult contact frequency, and twin resemblance was examined. Childhood closeness was established through questions that assessed whether twins shared the same room as children, shared the same playmates, dressed alike, and were in the same classes in school. Adult contact was measured by how often the twins saw each other and how often they were in phone or mail contact with each other. The 4 childhood items (4-point scale) were summed, as were the 2 adult variables (6-point scale). Pearson correlation coefficients were calculated between the contact variables and the absolute intrapair differences for the food pattern measures (the eating pattern score of twin 1 minus that of twin 2). A negative association indicated that twins with high levels of contact were more similar in their reported eating patterns than were those with lower levels of contact.

In addition, the sample was divided into groups with relatively low (<=12) and high (>12) scores on childhood closeness, as well as low (a few times a year or less) and high (once or twice a month or more) scores on adult contact, and monozygotic and dizygotic twin correlation coefficients for eating patterns were compared. Greater differences between monozygotic and dizygotic twins in the higher contact groups than in the lower contact groups would indicate violation of the equal environment assumption. Because sex differences could influence the results, opposite-sex twin pairs were excluded from this analysis.

Statistical analysis
Except for polyserial correlation analysis (35) and structural equation analysis (40), all analyses were performed with the statistical package SAS (43).


   RESULTS
TOP
 ABSTRACT
 INTRODUCTION
 SUBJECTS AND METHODS
 RESULTS
DISCUSSION
 APPENDIX A
 REFERENCES
 
Male twins were significantly older (x ± SD: 65.3 ± 7.8 y) than female twins (64.3 ± 8.0 y). Virtually all (>99%) participants were white. The men also had significantly higher mean weights (80.3 ± 12.4 compared with 65.1 ± 12.3 kg) and body mass indexes (in kg/m2: 25.6 ± 3.4 compared with 24.6 ± 4.4) than did the women. The sample is described in further detail in Table 1Go. Most subjects were living with a partner. Of subjects aged 50–65 y, 34% of women and 49% of men were retired compared with 71% and 83% of women and men aged >=66 y. Total family incomes >=$20000 were reported by {approx}79% of women and 89% of men aged 50–65 y and by 57% of women and 78% of men aged >=66 y. Most subjects had >=1–3 y of college education. Most subjects evaluated their health to be good to very good, although {approx}49% of women and 41% of men indicated they were consuming some type of special diet. The most frequently endorsed diets were low-salt and low-cholesterol diets. More than half of the sample used vitamin and mineral supplements.


View this table:
[in this window]
[in a new window]
  		TABLE 1. Sample descriptives1
 
Eating patterns
Factor analysis yielded 2 common factors that provided a straightforward and meaningful interpretation of the dietary data (Table 2Go). The first factor consisted of foods high in fat, salt, and sugar (a less healthful eating pattern, designated as eating pattern 1). The second factor included a variety of vegetables, fruit, rice, yogurt, skim milk, and dark bread (a healthful eating pattern, designated as eating pattern 2). Although factor analysis is usually based on independent observations, these factor scores were derived from a sample of related individuals. To test the extent to which this influenced the findings, we compared factor scores obtained from the entire sample with those derived from a sample of independent observations (eg, one member selected from each twin pair). Patterns were highly similar (CC: 0.99 for both factors; RMS deviation: 0.01 and 0.02 for factors 1 and 2, respectively), indicating the legitimacy of using both twin pair members in the factor analysis.


View this table:
[in this window]
[in a new window]
  		TABLE 2. Loadings of food-use items on factors 1 and 21
 
Similarly, we divided the sample into a male and a female group and tested whether sex differences in the loadings of food items on the 2 factors were apparent. Results indicated little difference between men and women (CC: 0.98 and 0.96 for factor 1 and 2, respectively; RMS deviation: 0.06 for both factors). Therefore, we used results from factor analysis based on the entire sample to obtain eating pattern scores for both men and women.

Scores on both eating patterns were calculated for each subject on the basis of the food items with the highest loadings (>=0.30) for the 2 factors. One food (corn bread) had factor loadings >0.30 with both factors 1 and 2 and therefore contributed to both eating patterns. Good agreement was found between the 2 factors obtained from factor analysis of food use items and the scored eating patterns (Pearson correlation coefficients of 0.96 for pattern 1 and 0.95 for pattern 2, respectively).

The fact that pattern 2 was associated with more healthful dietary habits than was pattern 1 was confirmed by the following polyserial correlations (use of pattern 1 and 2 foods, respectively): eating the skin on chicken, 0.34 and -0.03; eating the fat on meat, 0.34 and -0.05; and adding salt to food, 0.26 and -0.06. In addition, correlations with vitamin and mineral supplement intake were -0.04 and 0.16 (pattern 1 and 2 foods, respectively). The associations were comparable for eating patterns based on serving size and consumption frequency information.

Sex differences and stability of responses
Mean eating pattern scores for men and women separately are shown in Table 3Go. Mean scores for food use, serving size, and consumption frequency of pattern 1 foods were higher for men than for women. Men also ate larger portions of pattern 2 foods, but their mean use and consumption frequency scores for these foods were lower than those for women. Sex differences were significant for all 6 variables, P <= 0.05.


View this table:
[in this window]
[in a new window]
  		TABLE 3. Mean eating pattern scores1
 
For the subsample of 218 subjects who completed the questionnaire a second time, mean differences between the 2 test occasions were small and insignificant, with the exception of a significantly lower consumption frequency of pattern 2 foods for women on retesting. Test-retest correlation coefficients ranged from r = 0.68 (use of pattern 2 foods) to 0.86 (serving size of pattern 2 foods) for women and from 0.78 (use of pattern 1 foods) to 0.87 (serving size of pattern 1 foods) for men (all P <= 0.05).

Genetic analyses
Pearson correlation coefficients for pattern 1 and 2 food measures for the 5 twin-pair groups are shown in Figure 1Go. All correlation coefficients were significant (P <= 0.05), with the exception of pattern 1 and 2 serving size for dizygotic male and dizygotic opposite-sex twins, and pattern 2 use and consumption frequency for dizygotic male twins. Correlations for monozygotic twins were larger than those for dizygotic same-sex twins for all measures, suggesting that additive genetic influences contributed to all measures for both eating patterns. Differences were significant for pattern 2 food use and consumption frequency for men and pattern 1 and 2 food use and consumption frequency for women. For women, dizygotic correlation coefficients that were more than half of those for monozygotic twins contributed to all 3 measures of both eating patterns, suggesting the influence of common environmental factors. For men, evidence of common environmental influences was limited, with the possible exception of use and consumption frequency of pattern 1 foods. Comparison of correlation coefficients of opposite-sex twins with those of same-sex twins suggested that sex may have influenced twin resemblance for the eating patterns. Correlation coefficients for opposite-sex dizygotic twins were lower than those for both male and female dizygotic same-sex twins for food use and consumption frequency of pattern 1 foods and for serving size of pattern 2 foods. For the other variables, opposite-sex twin correlations were lower than those of same-sex female twins but higher than or equal to those for same-sex male twins. This difference was significant only for same-sex dizygotic female twins and dizygotic opposite-sex twins (P = 0.04).



View larger version (43K):
[in this window]
[in a new window]
  		FIGURE 1. Correlation coefficients for the twin groups. MZM, monozygotic male; DZM, dizygotic male; MZF, monozygotic female; DZF, dizygotic female; DZ-OS, dizygotic opposite-sex; consum freq, consumption frequency.

 
Results of the model-fitting are provided in Appendix A. General-effects sex-limitation models testing the possibility of different genes (or environments) for the sexes did not explain the data better than more parsimonious models (eg, the ACE model, the common-effects sex-limitation model, or the scalar sex-limitation model). Therefore, the results from these models are not discussed further. Sex-specific models explained the data significantly better for use and serving size of pattern 1 foods and for use of pattern 2 foods. For the remaining measures, no significant sex differences were found. Pie charts depicting relative genetic, common environmental, and specific environmental influences are presented in Figure 2Go.



View larger version (36K):
[in this window]
[in a new window]
  		FIGURE 2. Genetic and environmental influences on eating patterns 1 and 2. h2, heritable influences; c2, common environmental influences; e2, specific environmental influences.

 
The common-effects sex-limitation model with different magnitudes of genetic and environmental influences for men and women described the use of pattern 1 foods best (P = 0.02). The contribution of additive genetic variation to the total phenotypic variation was larger for men (0.38) than for women (0.30). These numbers represent the heritability estimates for the sexes. Common environmental influences contributed 0.17 to the total variation for women but were absent for men. The common-effects sex-limitation model also best explained the variation in serving size of pattern 1 foods (P = 0.04). The heritability estimate was larger for men (0.30) than for women (0.15). Common environmental influences explained 0.25 of the variation for women but did not contribute to the variation for men.

The ACE model with identical estimates for men and women provided the most parsimonious explanation of the variation in consumption frequency of pattern 1 foods. Approximately one-third (0.30) of the total variation was explained by heritable factors. Common environmental factors (<0.10) played a smaller role than specific environmental factors ({approx}0.60).

Use of pattern 2 foods was best explained by a scalar sex-limitation model that took sex differences in the total phenotypic variation into account (P = 0.01). The scalar value was estimated at 1.08 and accounted for the fact that the total phenotypic variation in pattern 2 food use was larger for men than for women. The relative contribution of genetic and environmental influences was identical for the sexes. Heritability was estimated to be 0.40, with specific environmental factors contributing an additional 0.48. The contribution of common environmental factors (0.12) was the smallest of the 3 influences.

Models with identical estimates for men and women provided the best explanation for both serving size and consumption frequency of pattern 2 foods. The relative contributions of heritable (0.33–0.39), common environmental (<0.10), and specific (>0.50) environmental influences were also similar for these 2 measures.

For all structural equation models presented in Figure 2, GoP values > 0.05 in combination with negative AIC values indicate good consistency between the models and the data (see Appendix A). CIs (95%) around the genetic estimates for the best-fitting models were as follows: pattern 1 food use, (0, 0.49) for men and (0.09, 0.52) for women; pattern 1 serving size, (0, 0.45) for men and (0, 0.45) for women; pattern 1 consumption frequency, (0.14, 0.46); pattern 2 food use, (0.24, 0.55); pattern 2 serving size, (0.16, 0.50); and pattern 2 consumption frequency, (0.14, 0.47). For common environmental influences the 95% CIs were as follows: pattern 1 food use, (0, 0.43) for men and (0, 0.35) for women; pattern 1 serving size, (0, 0.33) for men and (0, 0.44) for women; pattern 1 consumption frequency, (0, 0.24); pattern 2 food use, (0, 0.26); pattern 2 serving size, (0, 0.24); and pattern 2 consumption frequency, (0, 0.25). The most narrow CIs were found for specific environmental influences: pattern 1 food use, (0.51, 0.78) for men and (0.47, 0.59) for women; pattern 1 serving size, (0.55, 0.86) for men and (0.52, 0.69) for women; pattern 1 consumption frequency, (0.54, 0.66); pattern 2 food use, (0.44, 0.53); pattern 2 serving size, (0.50, 0.64); and pattern 2 consumption frequency, (0.53, 0.65).

Equal environment assumption
Monozygotic twins spent more time in each other's company than did dizygotic twins: 89% of the monozygotic twins but only 38% of the dizygotic twins reported that they were close in childhood. Also, 40% of monozygotic twins but only 21% of dizygotic twins reported having adult contact with their twin more than once a month. The possible influence of degree of closeness and contact on twin similarities in eating habits was examined by calculating correlations between childhood closeness and adult contact frequency and intrapair differences for the 6 eating pattern measures. Statistically significant negative associations were found for only 2 of 60 correlations (5 twin groups, 6 measures, 2 contact variables), both for monozygotic female twins: frequency of food consumption with childhood closeness (P = 0.04) and frequency of food consumption with adult contact frequency (P = 0.02). Thus, we found little evidence of an association between twin closeness or contact and resemblance for the eating patterns.

Twins were also divided into high and low childhood closeness groups and high and low adult contact groups and differences in correlation coefficients were compared for the monozygotic and dizygotic twin groups. Only same-sex twin pairs were included in the analysis. Of 24 comparisons (2 closeness-contact measures and 6 dietary measures for males and females), 14 suggested greater genetic involvement in the low rather than the high childhood closeness and adult contact categories. These findings provide additional evidence that the equal environment assumption was not violated in this study and that the results of the genetic analyses were not biased.


   DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 SUBJECTS AND METHODS
 RESULTS
 DISCUSSION
APPENDIX A
 REFERENCES
 
We identified 2 independent eating patterns, reflecting healthful and less healthful eating habits. Both genetic and environmental influences contributed to each eating pattern. Additive genetic influences contributed to about one-third of the total variation for most measures. The influence of the family environment was relatively small, whereas environmental influences not shared by family members explained most of the variation for the measures. Findings suggested sex differences in genetic and environmental contributions to several of the measures.

The relative magnitude of the additive genetic and environmental influences varied somewhat by eating pattern. Depending on the measure studied (ie, food use, serving size, or consumption frequency), genetic factors contributed between 15% and 38% to the total phenotypic variation in pattern 1 and between 33% and 40% in pattern 2. The contribution of common environmental influences was 0–25% for eating pattern 1 and 5–12% for eating pattern 2; specific environmental influences contributed 53–70% to eating pattern 1 and 48–58% to pattern 2.

Relation to other studies
This work confirms earlier studies showing that dietary information can be described in terms of a limited number of eating patterns, reflecting healthful and less healthful eating habits (1–7). Our findings also agree with previous twin studies indicating an influence of genes on eating behavior. Our estimates are generally higher than those reported by Pérusse et al (19) but lower than those reported by de Castro (20). Additionally, most of the estimates obtained by Heller et al (18) were lower than ours. Although Fabsitz et al (15) did not present heritability estimates, these can be derived from their correlation coefficients by using the formula 2 x ({rho}MZ - {rho}DZ) (44). Most of their estimates for individual food items were lower than or similar to ours, whereas most of their estimates for nutrients were higher.

This variation in findings between the current and previous studies may be the result of differences in methods. The present study differs from previous studies in the period covered by the dietary instrument [past year compared with 1 day or 1 week (15), 3 d (19), 4 d (18), or 7 d (20)] and in the type of measure used (eating patterns compared with individual food items or nutrients). The age range of the subjects in the present study was also more homogeneous than that in most other studies. Finally, previous studies did not take sex differences in genetic and environmental influences on diet into account.

Equal environment assumption
More frequent contact between monozygotic than dizygotic twins was found both in childhood and adulthood, confirming observations in previous studies (15, 45, 46). However, the association between childhood closeness and adult contact and twin resemblance for the eating patterns was negligible, making it unlikely that violation of the equal environment assumption inflated the heritability estimates in this study.

Sex differences
The dietary patterns identified with factor analysis were similar for men and women. However, significant mean sex differences for all 3 measures of both eating patterns reflected healthful eating patterns in women and larger portion sizes for men, confirming previous reports (26, 47–49).

Structural equation modeling suggested sex-specificity in the contributions of genetic and environmental influences to 3 of the 6 measures studied. The model with different magnitudes of genetic and environmental influences for the sexes (common-effects sex-limitation model) best explained food use and serving size for eating pattern 1. Genetic influences were larger for men, whereas common environmental influences were larger for women. In addition, a model allowing sex differences in total phenotypic variation best explained the use of pattern 2 foods. Models specifying different sets of genetic factors or different environmental factors (general-effects sex-limitation models) did not fit the data better than the other models tested. Therefore, it appears that sex differences in eating patterns are the result of the same rather than different sets of genes (or environmental factors), but that the magnitude of these influences may differ between the sexes.

Our finding of greater common environmental influences for women for some measures may reflect environmental pressures to limit intakes of certain foods. For example, women are more likely than men to consume weight-reducing diets. In the present study, 12.6% of women, compared with 7.5% of men, indicated adherence to a weight-loss diet. Strong environmental forces can reduce the estimated contribution of genetic influences (50). Thus, genetic estimates for women may be more similar to those for men when measured in an environment that places equal pressures on both sexes.

Statistical considerations
Because of differences in sample size, power was greatest for analyses of food use and lowest for analyses of serving size. In addition, power was greater for analyses based on the full sample (basic ACE model) than on models that considered male and female samples separately. Compared with additive genetic and specific environmental factors, significant common environmental influences are more difficult to detect with a classic twin design. This is especially the case when these influences are modest in size (as our results indicate). Even though our study used the largest twin sample to date to study food intake, sample sizes several times larger would have been required to detect significant common environmental influences (51).

Results were based on the full model, including additive genetic, common environmental, and specific environmental influences. CIs (95%) were larger for genetic and common environmental influences than for specific environmental influences. CIs around genetic estimates included zero for use of pattern 1 foods in men and for serving size of pattern 1 foods in men and women. In addition, all CIs around common environmental estimates included zero, whereas none for specific environmental factors did.

Study limitations
The sample was of middle to older age, predominantly female, white, and of relatively high socioeconomic status, and may not be representative of the general population. Additionally, food-frequency questionnaires rely on self-reported intake, which may be biased (52, 53). However, even though self-reported measures may not reflect absolute food intake, they are useful for ranking individuals according to their intakes or for comparing the intakes of subgroups (27). Therefore, this type of questionnaire was judged sufficient for the present study. Other limitations of this study were that diet was assessed over the past year, which may not be representative of the subjects' lifetime intakes. In addition, self-reports of dietary intake may include error (9). The mail-in dietary questionnaires had more missing values than might have resulted if a face-to-face interview with probing questions had been used. However, in twin research, relatively large sample sizes are required and the tradeoff of increased inaccuracy for larger numbers of respondents was considered necessary.

Conclusion
More and less healthful dietary patterns were identified in twins aged >=50 y. Inherited biological differences played a role in explaining these patterns, although the effect of environmental factors was larger. Environmental influences not shared by family members had a greater effect than did familial environmental influences. For half of the measures studied, genetic and environmental influences differed depending on sex. In general, men were somewhat more likely to exhibit genetic effects than were women. These findings may be important in crafting dietary interventions and in predicting dietary adherence.


   APPENDIX A
TOP
 ABSTRACT
 INTRODUCTION
 SUBJECTS AND METHODS
 RESULTS
 DISCUSSION
 APPENDIX A
REFERENCES
 


View this table:
[in this window]
[in a new window]
  		Results of model-fitting1
 

   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 SUBJECTS AND METHODS
 RESULTS
 DISCUSSION
 APPENDIX A
 REFERENCES
 
  1. Schwerin HS, Stanton JL, Riley AM Jr, et al. Food eating patterns and health: a reexamination of the Ten-State and HANES 1 surveys. Am J Clin Nutr 1981;34:568–80.[Abstract/Free Full Text]
  2. Schwerin HS, Stanton JL, Smith JL, Riley AM Jr, Brett BE. Food, eating habits, and health: a further examination of the relationship between food eating patterns and nutritional health. Am J Clin Nutr 1982;35:1319–25.[Free Full Text]
  3. Akin JS, Guilkey DK, Popkin BM, Fanelli MT. Cluster analysis of food consumption patterns of older Americans. J Am Diet Assoc 1986;86:616–24.[Medline]
  4. Hulshof KFAM, Wedel M, Löwik MRH, et al. Clustering of dietary variables and other lifestyle factors (Dutch Nutritional Surveillance System). J Epidemiol Community Health 1992;46:417–24.[Abstract]
  5. Randall E, Marshall JR, Graham S, Brasure J. Patterns in food use and their associations with nutrient intakes. Am J Clin Nutr 1990;52:739–45.[Abstract/Free Full Text]
  6. Tucker KL, Dallal GE, Rush D. Dietary patterns of elderly Boston-area residents defined by cluster analysis. J Am Diet Assoc 1992;92:1487–91.[Medline]
  7. Huijbregts PPC, Feskens EJM, Kromhout D. Dietary patterns and cardiovascular risk factors in elderly men: the Zutphen Elderly Study. Int J Epidemiol 1995;24:313–20.[Abstract/Free Full Text]
  8. Kumanyika S, Tell GS, Shemanski L, Polak J, Savage PJ. Eating patterns of community-dwelling older adults: the Cardiovascular Health Study. Ann Epidemiol 1994;4:404–15.[Medline]
  9. Birch LL. The relationship between children's food preferences and those of their parents. J Nutr Educ 1980;12:14–8.
  10. Pliner P, Pelchat ML. Similarities in food preferences between children and their siblings and parents. Appetite 1986;7:333–42.[Medline]
  11. Logue AW, Logue CM, Uzzo RG, McCarty MJ, Smith ME. Food preferences in families. Appetite 1988;10:169–80.[Medline]
  12. Rozin P. Family resemblance in food and other domains: the family paradox and the role of parental congruence. Appetite 1991; 16:93–102.[Medline]
  13. Oliveria SA, Ellison RC, Moore LL, Gillman MW, Garrahie EJ, Singer MR. Parent-child relationships in nutrient intake: the Framingham Children's Study. Am J Clin Nutr 1992;56:593–8.[Abstract/Free Full Text]
  14. Stafleu A, Van Staveren WA, de Graaf C, Burema J, Hautvast JGAJ. Family resemblance in energy, fat, and cholesterol intake: a study among three generations of women. Prev Med 1994;23:474–80.[Medline]
  15. Fabsitz RR, Garrison RJ, 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;8:15–24.[Medline]
  16. Wade J, Milner J, Krondl M. Evidence for a physiological regulation of food selection and nutrient intake in twins. Am J Clin Nutr 1981;34:143–7.[Abstract/Free Full Text]
  17. Krondl M, Coleman P, Wade J, Milner J. A twin study examining the genetic influence on food selection. Hum Nutr Appl Nutr 1983; 37:189–98.
  18. Heller RF, O'Connell DL, Roberts DCK, et al. Lifestyle factors in monozygotic and dizygotic twins. Genet Epidemiol 1988;5:311–21.[Medline]
  19. Pérusse L, Tremblay A, Leblanc C, et al. Familial resemblance in energy intake: contribution of genetic and environmental factors. Am J Clin Nutr 1988;47:629–35.[Abstract/Free Full Text]
  20. de Castro JM. Genetic influences on daily intake and meal patterns of humans. Physiol Behav 1993;53:777–82.[Medline]
  21. Falciglia GA, Norton PA. Evidence for a genetic influence on preference for some foods. J Am Diet Assoc 1994;94:154–8.[Medline]
  22. Faust J. A twin study of personal preferences. J Biosoc Sci 1974; 6:75–91.[Medline]
  23. Rozin P, Millman L. Family environment, not heredity, accounts for family resemblances in food preferences and attitudes: a twin study. Appetite 1987;8:125–34.[Medline]
  24. Kendler KS, Neale MC, Kessler RC, Heath AC, Eaves LJ. A test of the equal-environment assumption in twin studies of psychiatric illness. Behav Genet 1993;23:21–7.[Medline]
  25. Lykken DT, McGue M, Bouchard TJ Jr, Tellegen A. Behav Genet 1990;20:547–61.[Medline]
  26. Block G, Hartman AM, Dresser CM, Carroll MD, Gannon J, Gardner L. A data-based approach to diet questionnaire design and testing. Am J Epidemiol 1986;124:453–69.[Abstract/Free Full Text]
  27. Mares-Perlman JA, Klein BEK, Klein R, Ritter LL, Freudenheim JL, Luby MH. Nutrient supplements contribute to the dietary intake of middle- and older-aged adult residents of Beaver Dam, Wisconsin. J Nutr 1993;123:176–88.
  28. Sobell J, Block G, Koslowe P, Tobin J, Andres R. Validation of a retrospective questionnaire assessing diet 10–15 years ago. Am J Epidemiol 1989;130:173–87.[Abstract/Free Full Text]
  29. Cummings SR, Block G, McHenry K, Baron RB. Evaluation of two food frequency methods of measuring dietary calcium intake. Am J Epidemiol 1987;126:796–802.[Abstract/Free Full Text]
  30. Block G, Woods MN, Potosky A, Clifford CA. Validation of a self-administered diet history questionnaire using multiple diet records. J Clin Epidemiol 1990;43:1327–35.[Medline]
  31. Cohen DJ, Dibble E, Grawe JM, Pollin W. Reliably separating identical from fraternal twins. Arch Gen Psychiatry 1975;32:1371–5.[Abstract]
  32. Cederlöf R, Friberg L, Jonsson E, Kaij L. Studies on similarity diagnosis in twins with the aid of mailed questionnaires. Acta Genet Stat Med 1961;11:338–62.[Medline]
  33. Gorsuch RL. Factor analysis. 2nd ed. Hillsdale, NJ: Lawrence Erlbaum Associates, 1996.
  34. Harman HH. Modern factor analysis. 3rd ed. Chicago: The University of Chicago Press, 1976.
  35. Jöreskog KG, Sörbom D. PRELIS 2 user's reference guide. Chicago: Scientific Software Inc, 1993.
  36. Neale MC, Cardon LR. Methodology for genetic studies of twins and families. Dordrecht, Netherlands: Kluwer Academic Publishers, 1992.
  37. Heath AC, Neale MC, Hewitt JK, Eaves LJ, Fulker DW. Testing structural equation models for twin data using LISREL. Behav Genet 1989;19:9–35.[Medline]
  38. Blom G. Statistical estimates and transformed beta variables. New York: Wiley and Son, 1958.
  39. Snedecor GW, Cochran WG. Statistical methods. 6th ed. Ames, IA: Iowa State University Press, 1979.
  40. Neale MC. Mx: statistical modelling. 3rd ed. Richmond, VA: Department of Psychiatry, Medical College of Virginia, Virginia Commonwealth University, 1995.
  41. Eaves LJ, Krystyna A, Last P, Young A, Martin NG. Model-fitting approaches to the analysis of human behaviour. Heredity 1978; 41:249–320.[Medline]
  42. Akaike H. Factor analysis and AIC. Psychometrika 1987;52:317–32.
  43. SAS Institute Inc. SAS procedures guide. Version 6, 3rd ed. Carey, NC: SAS Institute Inc, 1990.
  44. Plomin R, DeFries JC, McClearn GE. Behavioral genetics: a primer. San Fransisco: WH Freedman and Company, 1980.
  45. Loehlin JC, Nichols RC. Heredity, environment and personality: a study of 850 sets of twins. Austin, TX: University of Texas Press, 1976.
  46. Rose RJ, Kaprio J, Williams CJ, Viken R, Obremski K. Social contact and sibling similarity: facts, issues, and red herrings. Behav Genet 1990;20:763–78.[Medline]
  47. Cronin FJ, Krebs-Smith SM, Wyse BW, Light L. Characterizing food usage by demographic variables. J Am Diet Assoc 1982;81:661–73.[Medline]
  48. Logue AW, Smith ME. Predictors of food preferences in adult humans. Appetite 1986;7:109–25.[Medline]
  49. Verbrugge LM. Sex differentials in health. Public Health Rep 1982;97:417–37.[Medline]
  50. Heath AC, Berg K, Eaves LJ, et al. Education policy and the heritability of educational attainment. Nature 1985;314:734–6.[Medline]
  51. Martin NG, Eaves LJ, Kearsey MJ, Davies P. The power of the classical twin study. Heredity 1978;40:97–116.[Medline]
  52. Myers RJ, Klesges RC, Eck LH, Hanson CL, Klem ML. Accuracy of self-reports of food intake in obese and normal-weight individuals: effects of obesity on self-reports of dietary intake in adult females. Am J Clin Nutr 1988;48:1248–51.[Abstract/Free Full Text]
  53. Mertz W, Tsui JC, Judd JT, et al. What are people really eating? The relation between energy intake derived from estimated diet records and intake determined to maintain body weight. Am J Clin Nutr 1991;54:291–5.[Abstract/Free Full Text]
Received for publication August 5, 1998. Accepted for publication March 25, 1999.




This article has been cited by other articles:


Home page
 Am. J. Clin. Nutr.Home page
K. Keskitalo, H. Tuorila, T. D Spector, L. F Cherkas, A. Knaapila, J. Kaprio, K. Silventoinen, and M. Perola
The Three-Factor Eating Questionnaire, body mass index, and responses to sweet and salty fatty foods: a twin study of genetic and environmental associations
Am. J. Clinical Nutrition, August 1, 2008; 88(2): 263 - 271.
[Abstract] [Full Text] [PDF]

Home page
 Am. J. Clin. Nutr.Home page
M. S Faith, S. A. Rhea, R. P Corley, and J. K Hewitt
Genetic and shared environmental influences on children's 24-h food and beverage intake: sex differences at age 7 y
Am. J. Clinical Nutrition, April 1, 2008; 87(4): 903 - 911.
[Abstract] [Full Text] [PDF]

Home page
 Am. J. Clin. Nutr.Home page
K. Keskitalo, H. Tuorila, T. D Spector, L. F Cherkas, A. Knaapila, K. Silventoinen, and M. Perola
Same genetic components underlie different measures of sweet taste preference
Am. J. Clinical Nutrition, December 1, 2007; 86(6): 1663 - 1669.
[Abstract] [Full Text] [PDF]

Home page
 DiabetesHome page
M. de Krom, Y. T. van der Schouw, J. Hendriks, R. A. Ophoff, C. H. van Gils, R. P. Stolk, D. E. Grobbee, and R. Adan
Common Genetic Variations in CCK, Leptin, and Leptin Receptor Genes Are Associated With Specific Human Eating Patterns
Diabetes, January 1, 2007; 56(1): 276 - 280.
[Abstract] [Full Text] [PDF]

Home page
 Am. J. Clin. Nutr.Home page
S. Tholin, F. Rasmussen, P. Tynelius, and J. Karlsson
Genetic and environmental influences on eating behavior: the Swedish Young Male Twins Study
Am. J. Clinical Nutrition, March 1, 2005; 81(3): 564 - 569.
[Abstract] [Full Text] [PDF]

Home page
 Am. J. Clin. Nutr.Home page
G. Cai, S. A Cole, R. A Bastarrachea-Sosa, J. W MacCluer, J. Blangero, and A. G Comuzzie
Quantitative trait locus determining dietary macronutrient intakes is located on human chromosome 2p22
Am. J. Clinical Nutrition, November 1, 2004; 80(5): 1410 - 1414.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Purchase Article
Right arrow View Shopping Cart
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by van den Bree, M. B.
Right arrow Articles by Dwyer, J. T
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by van den Bree, M. B.
Right arrow Articles by Dwyer, J. T
Agricola
Right arrow Articles by van den Bree, M. B.
Right arrow Articles by Dwyer, J. T

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS