• twins;
  • exercise;
  • obesity/epidemiology;
  • weight gain/genetics;
  • National Longitudinal Study of Adolescent Health


  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgement
  8. References

Objective: The magnitude of environmental vs. genetic effects on BMI, diet, and physical activity (PA) is widely debated. We followed a sibling cohort (where individuals shared households in childhood and adolescence) to young adulthood (when some continued sharing households and others lived apart) to examine the role of discordant environments in adult twins’ divergent trends in BMI and health behaviors and to quantify the variation in BMI and behavior among all siblings that is attributable to environmental and additive genetic effects.

Research Methods and Procedures: In the National Longitudinal Study of Adolescent Health, siblings sharing households for ≥10 years as adolescents (mean age = 16.5 ± 1.7 years; N = 5524) were followed into adulthood (mean = 22.4 ± 1.8 years; N = 4368), self-reporting PA, sedentary behavior, and dietary characteristics. Adult BMI and adolescent z scores were derived from measured height and weight.

Results: Compared with those living together, twins living apart exhibited greater discordance in change in BMI, PA, and fast food intake from adolescence to adulthood. Adolescent household environments accounted for 8% to 10% of variation in adolescent fast food intake and sedentary behaviors and 50% of variation in adolescent overweight. Adolescent household effects on PA were substantially greater in young adulthood (accounting for 50% of variation) vs. adolescence. Young adult fast food intake was significantly affected by young adult household environment, accounting for 12% of variation.

Discussion: These findings highlight important environmental influences on BMI, PA, and fast food intake during the transition to adulthood. Household and physical environments play an important role in establishing long-term behavior patterns.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgement
  8. References

The transition from childhood to adulthood is a critical stage, marked by striking physical activity (PA)1 declines (1) and weight gain (2). Household and neighborhood environments have substantial effects on body weight (e.g., (3, 4), related health behaviors [e.g., PA (5, 6), and fast food intake (7)] and are important modifiable risk factors for obesity prevention/reduction strategies.

Quantifying genetic and environmental influences on health outcomes is challenging. Genetic factors affect physiological processes related to PA (8) and weight (9, 10), but epidemiological studies of environment effects typically do not account for genetic predisposition. The extent to which genes contribute to body size, exercise, and/or food preferences is widely debated. For example, an estimated 50% to 90% of variability in adiposity is attributable to genetics (11). Population-wide genetic differences may substantially limit the ability to detect environmental effects on health (particularly modest effects) using cohort studies of non-related individuals.

Many study designs that are used to address genetic variation, including studies of twins separated at birth and reared apart (e.g., (12), do not allow separation of effects of different environment exposures (e.g., separation of childhood influences from later environmental effects). Family/household environments during childhood may influence long-term health behaviors and may confound or bias estimates of environment-health associations. These methodological concerns continue to fuel the widespread debate over the true magnitude of environmental effects on obesity and related health behaviors.

Pedigree studies provide a unique opportunity to address genetic variation in a population and the influence of unmeasured or poorly measured confounding encountered in observational cohort studies. The transition from adolescence to adulthood, when a substantial number of individuals relocate to new environments, allows assessment of relative contributions of environmental and genetic influences in shared vs. non-shared environments. Following individuals who are genetically similar and share childhood household and neighborhood environments, we can evaluate variation in health behaviors and subsequent behavior changes as individuals move into different environments. By concurrently assessing individuals with varying degrees of genetic relatedness (in shared households as children and adolescents), we can account for multiple shared environmental and genetic sources of variation, providing the opportunity to better tease apart causal effects on health using methods not possible with traditional cohort studies.

Our analyses use sibling data from the National Longitudinal Study of Adolescent Health (Add Health). As adolescents, participants shared household and neighborhood environments plus various other unmeasured genetic and familial factors. As young adults, some continued to share households, whereas others moved to new environments. Therefore, this design provides a natural experiment that can be used to supplement findings from cohort studies quantifying environmental effects on health trends during this critical stage of life. The specific aims of this paper were to examine the correlations of various health measures in monozygotic (MZ) and same-sex dizygotic (DZ) twins continuing to live together as young adults, vs. those who live apart, and to quantify additive genetic and household effects on BMI and related health behaviors in a large sample of siblings and non-siblings sharing family and household environments.

Research Methods and Procedures

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgement
  8. References

Add Health

Add Health is a nationally representative, longitudinal survey of youths, grades 7 to 12. Survey procedures described elsewhere (13) were approved by the Institutional Review Board, University of North Carolina at Chapel Hill. Wave I (1994 to 1995) included 20, 745 adolescents and their parents. Wave II (1996, n = 14, 438) included Wave I adolescents who had not graduated from high school, including drop-outs. Wave III (2001 to 2002, n = 15, 197) included all located Wave I respondents, now 18 to 27 years old (regardless of Wave II participation).

Family Subsample

Add Health oversampled subgroups, including a respondent subsample matched with related and non-related adolescents sharing a Wave I household. Up to five adolescents per household were surveyed (mean, 2.1 ± 0.4 individuals/household), including twins, full siblings, one-half siblings, cousins, and non-related adolescents (Figure 1). Cousins were not assessed by degree of relatedness. The sample included 5524 Wave I respondents living in 2639 households. We limited our sample to 4782 adolescents (in 2302 households) living together for ≥10 years at Wave I, thus sharing household/physical environments during childhood and adolescence. In Wave III, 4588 participants were surveyed as young adults; 913 individuals (in 449 households) continued living with another survey respondent. We then excluded participants who were severely disabled (n = 100), pregnant (Wave II, n = 75; III, n = 94), or missing primary exposure variables (BMI, Wave II, n = 450; III, n = 613; PA, Wave II, n = 437; III, n = 532; sedentary behaviors, Wave II, n = 436; III, n = 567; fast food, Wave II: n = 440; III, n = 515).


Figure 1. Shared and non-shared environments among the sibling and household analysis subsample.

Download figure to PowerPoint

Non-twin sibships were classified by self-report. Twin zygosity was determined by matching 12 molecular genetic markers at Waves I or III (n = 808) (14) or by full agreement of self-report measures, including confusability of appearance (n = 683) (http:www.cpc.unc.eduprojectsaddhealth). Zygosity was not determined for 83 participants who were excluded. Buccal cells were collected for DNA analysis by rubbing cheeks/gums with a sterile cytology brush. DNA extraction, conducted at the University of Arizona using established methods (15, 16, 17), yielded, on average, 58 ± 1 μg DNA/sample. DNA was stored at 4 °C and transported for genotyping to the Institute of Behavioral Genetics, University of Colorado (Boulder, CO). Questionable results were repeated to limit potential genotyping errors.


BMI (kilograms per meter squared) was calculated from measured height and weight, assessed at Waves II and III using standardized procedures. Self-reported height and weight were substituted for those refusing measurement and/or weighing more than the scale capacity (Wave II, n = 54; III, n = 155) (1). Wave II BMI z scores were based on National Center for Health Statistics/Centers for Disease Control and Prevention reference data (18). Adolescent overweight was defined as BMI ≥ 95th percentile (19) [individuals aged >20 years (n = 14) were considered 20 years of age]. In adulthood (Wave III), obesity was classified as BMI ≥ 30 (20).

Daily Activities

Daily PA (e.g., housework, hobbies, active play, sports, exercise) was assessed using standard 7-day recall questionnaire methods relevant for epidemiological research (21). Described in detail elsewhere (5, 22), questions were similar to those used and validated in other large-scale studies (21, 23, 24) and allowed calculation of activity frequency (bouts per week) by metabolic equivalent level. Moderate-vigorous (MV) PA was 5 to 8 metabolic equivalents (25). Adolescents reported sedentary behaviors (hours per week watching TV/videos and/or playing video/computer games). Wave III included questions appropriate for young adults (e.g., walking for exercise, weight-lifting). As in previous Add Health literature (2), a scaled MVPA sum was used, corresponding to the number of activities reported in Waves I and II to account for potential reporting bias and artificial increase in reported activities due to inclusion of additional questionnaire items. MVPA frequency was summed to determine whether participants met national PA recommendations (≥5 bouts/wk) (26).

Dietary habits (e.g., times/wk eating at a fast food restaurant) were measured using a 7-day recall instrument. Changes in activity and diet were assessed as the difference between values reported at Waves II and III.


Aim 1

Twin pair correlations were used to assess whether discordant environments were associated with differential changes in BMI and health behaviors. Twin correlations were determined by Pearson correlation coefficients, stratified by zygosity (STATA version 8.2, 2004; STATA Corporation, College Station, TX). In Aim 1, only MZ and same-sex DZ twins were assessed. Cross-sectional (Waves II and III) and longitudinal (Waves II to III) analyses were conducted. Analyses were stratified by household residence in adulthood (living together/apart in Wave III).

The expectation was that traits determined predominantly by genetic vs. environmental factors would yield MZ pair correlations two times that of DZ pair correlations; however, for traits more environmentally determined, twin correlations would be higher among those in shared (vs. non-shared) households, regardless of zygosity.

The difference between correlation coefficients in pairs living together vs. apart, was tested using Fisher's z score transformation, computing a z score for each coefficient. z Score differences were divided by the standard error of difference between the correlations (Equation 1).

  • image

To determine statistical significance of the difference in correlations, the resulting z value was compared with a value of 1.96 (i.e., α = 0.05) (27).

Aim 2

Using a variance component approach, SOLAR software (28) was used to estimate the proportion of observed variation in a given health outcome attributable to additive genetic effects, household/physical environment, and other unidentified sources (i.e., residual effects). Heritability is the ratio of additive genetic variance (σ2g) to total phenotypic variancey (σ2p) (Equation 2).

  • image

Individual phenotype (y) was modeled as a linear function accounting for genetic (g), household (c), and environmental (e) effects and potentially important covariates (age, sex) (Equation 3). Analyses assumed that gi, ci, and ei were normally distributed with a mean of 0 and variance of σ2g, σ2h, and σ2c, respectively.

  • image

Phenotypic variation within a set of relative pairs (ω) was estimated as a function of additive genetic effects (2Ψσ2g), shared household effects (Cσ2c), and residual effects (Iσ2e) (Equation 4). These parameters represent the estimated percentage of variation in a health outcome (for a related pair of individuals) due to additive genetic, environmental, and residual effects, respectively.

  • image

Accounting for varying pedigree structures, each of the variances is multiplied by a structuring matrix. The structuring matrix for additive genetic variance is two times the matrix of kinship coefficients (Ψ); for environmental variance it is an identity matrix (I) (diagonal elements are ones, remaining elements are zero), and for household variance it is a household matrix (C), which permits unique environments for each household. Household environment effects were estimated using shared or non-shared residence to approximate the household.

We employed a bivariate variance component approach to assess the joint effect of genes, households, and environment on the covariation of BMI and related behaviors. In multivariate models, phenotype covariance was decomposed into genetic, household, and environmental components (Equation 5),

  • image

where h21 and h22, c21 and c22, and e21 and e22, are heritability, additive household variance, and environmental variance of traits 1 and 2, respectively. ρg, ρe, and ρc are additive genetic, household, and individual-specific environmental correlations between the two traits, respectively.

Given the expected mean and covariance matrix, a multivariate normal density function was used to estimate the likelihood of household members’ phenotypes. Likelihoods were summed for all individuals. For variables with kurtotic distributions, we employed the multivariate Student's t distribution because simulation studies have shown that it recovers the correct Type I error, even in cases of strong kurtosis (29).


  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgement
  8. References

Mean age was 16.5 ± 1.7 years in Wave II and 22.4 ± 1.8 years in Wave III. At baseline, 50.3% of the sample was female. Prevalence of overweight/obesity increased from adolescence to adulthood; MVPA and breakfast frequency declined (Table 1). There was little change in sedentary behavior and fast food intake.

Table 1.  Overweight/obesity, MVPA, and dietary characteristics among adolescents (Wave II) and young adults (Wave III)
 AdolescenceYoung adulthood
  • MVPA, moderate-vigorous physical activity; SD, standard deviation.

  • *

    ≥95th percentile BMI-for-age z-score (18).

  • BMI ≥ 30 (kg/m2).

BMI (mean kg/m2 ± SD)22.9 (±4.9)26.2 (±6.2)
Percentage overweight/obese11.3%*21.6%
MVPA (mean bouts/wk ± SD)3.5 (±2.1)1.6 (±2.0)
Percentage with ≥5 bouts of MVPA per week29.4%8.6%
Sedentary behavior (mean h/wk ± SD)23.1 (±22.7)23.6 (±21.6)
Fast food intake (mean times/wk ± SD)2.2 (±1.8)2.5 (±2.1)
Breakfast intake (mean d/wk ± SD)4.2 (±2.6)2.9 (±2.7)

Aim 1: Twin Correlations

Wave II Cross-Sectional Findings

Compared with Wave II same-sex DZ twins, twin pair correlations were greater among MZ twins for BMI z score (MZ and DZ, 0.74 and 0.44, respectively), total weekly MVPA (0.56 and 0.41), weekly fast food intake (0.52 and 0.39), and weekly breakfast consumption (0.43 and 0.33). Correlations for total weekly hours of sedentary behavior followed opposite trends (0.32 and 0.40).

Waves II to III Longitudinal Findings

Table 2 shows correlations in health measures by residence (living together or apart). Twins living apart had greater discordant longitudinal changes in BMI, MVPA, sedentary behavior (DZ twins only), fast food consumption, and breakfast consumption (MZ twins only). However, differences in correlations were statistically significant only for BMI gain (MZ twins) and MVPA change (DZ twins). Conversely, MZ twins living apart were significantly more highly correlated in sedentary behavior change.

Table 2.  MZ and same-sex DZ twin pair correlations by shared and non-shared household environments: longitudinal change in health indicators from adolescence to young adulthood (Wave III-Wave II)
 MZ correlation coefficient (N)DZ correlation coefficient (N)
 Live togetherLive apartLive togetherLive apart
  • MZ, monozygotic; DZ, dizygotic; MVPA, moderate-vigorous physical activity.

  • *

    Correlation significantly different from zero (p < 0.05).

  • Correlation of those living together significantly differs from that of those living apart (p < 0.05).

Change in BMI (kg/m2)0.63 (142)*0.40 (214)*,0.36 (88)*0.21 (200)*
Change in MVPA (bouts/wk)0.30 (150)*0.17 (226)*0.40 (90)*0.12 (200)
Change in sedentary behavior (h/wk)−0.06 (148)0.31 (220)*,0.31 (88)*0.18 (198)*
Change in fast food (times/wk)0.21 (148)*0.17 (226)*0.38 (90)*0.32 (202)*
Change in breakfast (d/wk)0.33 (148)*0.19 (226)*0.12 (90)0.27 (202)*
Wave III Cross-Sectional Findings

MZ twins living apart had significantly less similarity in BMI at Wave III than MZ twins living together (Table 3). DZ twins living apart were less similar in fast food intake at Wave III than DZ twins living together. MZ twins living apart were more highly correlated in sedentary behavior compared with those living together, and DZ twins living apart were more similar in achieving ≥5 bouts MVPA per week.

Table 3. . MZ and same-sex DZ twin pair correlations by shared and non-shared household environments: cross-sectional health indicators in young adulthood (Wave III).
 MZ correlation coefficient (N)DZ correlation coefficient (N)
 Live togetherLive apartLive togetherLive apart
  • MZ, monozygotic; DZ, dizygotic; MVPA, moderate-vigorous physical activity.

  • *

    Correlation significantly different from zero (p < 0.05).

  • Correlations of those living together significantly differs from that of those living apart (p < 0.05).

BMI (kg/m2)0.91 (154)*0.81 (234)*,0.25 (94)*0.31 (222)*
BMI ≥30 (kg/m2)0.70 (154)*0.66 (234)*0.04 (94)0.22 (222)*
MVPA (bouts/wk)0.25 (162)*0.35 (246)*0.22 (96)*0.18 (226)*
≥5 bouts MVPA/wk0.28 (162)*0.28 (242)*−0.13 (96)0.14 (218)*,
Sedentary behavior (h/wk)0.16 (160)*0.40 (240)*,0.16 (94)0.09 (224)
Fast food (times/wk)0.29 (162)*0.18 (246)*0.40 (96)*0.11 (228)
Breakfast (d/wk)0.44 (162)*0.32 (246)*0.17 (96)0.22 (228)*

Aim 2: Sibling/Household Variance Effect Estimates

Phenotypic Variance Estimates

Shared adolescent households influenced BMI and related health behaviors in adolescence and young adulthood (Table 4). The proportion of variance attributable to genes ranged from 0.30 to 0.40 on adolescent health measures, with the exception of ≥5 weekly bouts MVPA (0.59), and was significant for all measures. Household effects were significant for all but MVPA, ranging from 0.08 to 0.12 for dietary and sedentary behaviors and 0.51 for overweight (≥95th percentile BMI). Approximately one-half of the variation in diet and sedentary behavior was not accounted for.

Table 4. . Phenotypic variance in weight-related health measures in adolescence and adulthood due to genes, adolescent household environment, and residual factors
 Proportion variance due to additive genetic effects (SE)Proportion variance due to adolescent household environment effects (SE)Proportion variance due to residual effects
  • SE, standard error; MVPA, moderate-vigorous physical activity.

  • *

    Variance effects, p < 0.05.

  • Age effects, p < 0.05.

  • Sex effects, p < 0.05.

Adolescent health measures   
 ≥95th percentile BMI (n = 4794)0.32 (0.08)*0.51 (0.06)*0.17
 ≥5 bouts MVPA/wk (n = 4809)†,0.59 (0.08)*0.02 (0.07)0.39
 Sedentary behavior (h/wk) (n = 4755)†,0.34 (0.08)*0.10 (0.04)*0.56
 Fast food (times/wk) (n = 4806)†,0.34 (0.08)*0.08 (0.05)*0.58
 Breakfast (d/wk) (n = 4807)†,0.33 (0.07)*0.12 (0.04)*0.55
Young adult health measures   
 BMI ≥ 30 (n = 3787)†,0.54 (0.12)*0.28 (0.10)*0.18
 ≥5 bouts MVPA/wk (n = 3868)0.12 (0.18)0.50 (0.11)*0.38
 Sedentary behavior (h/wk) (n = 3833)†,0.28 (0.11)*0.04 (0.06)0.68
 Fast food (times/wk) (n = 3883)†,0.24 (0.09)*0.03 (0.05)0.73
 Breakfast (d/wk) (n = 3883)0.27 (0.10)*0.06 (0.06)0.67

The magnitude of total genetic and adolescent household effects on young adult obesity (BMI ≥ 30 kg/m2) was similar to that observed in adolescence, ∼0.83 (Table 4). Adolescent household effects on MVPA, however, were substantially greater in young adulthood (0.50). Household effects on diet and sedentary behavior ranged from 0.03 to 0.06, and genetic effects ranged from 0.24 to 0.28. Residual variation in adult estimates was similar in magnitude to that during adolescence.

We observed few significant joint influences of genes, households, or residual effects on BMI and other characteristics (not shown). The additive genetic correlations between traits were statistically significant (p < 0.05) for Wave II BMI z score and television viewing (0.10), Wave II BMI z score and fast food intake (−0.24), and Wave II BMI z score and breakfast consumption (−0.20). Additive genetic correlations between traits were also significant for Wave III BMI and breakfast consumption (−0.19). In addition, correlation of residual effects was significant for Wave III BMI and television viewing (0.21) and Wave III BMI and total sedentary behavior hours (0.19).

Differential Effects of Environment

Adolescent household effects were more influential in determining extreme BMI levels (Figure 2). The proportion of variance attributable to household effects in adolescence was substantially greater in at-risk-for-overweight (85th percentile) and overweight (95th percentile) than in the full BMI z score range.


Figure 2. Proportion of variance in Wave II adolescent weight status (assessed by various measures) due to additive gene, household environment, and residual effects.

Download figure to PowerPoint

Young Adult Household Effects

We expected environmental effects in young adulthood to represent the truest effect of household and physical environments, independently of other confounding characteristics of the family, parental attributes, and childhood up-bringing. Our results, however, indicated that shared young adult households accounted for little variation in adult BMI and health behaviors, although genetic effects were still prominent. Notably, a significant Wave III household effect was detected for fast food consumption (0.12 ± 0.02), accounting for ∼12% of variation. Variance attributable to additive genetic effects was significant for BMI ≥ 30 (0.85 ± 0.03), ≥5 bouts MVPA per week (0.69 ± 0.07), sedentary behavior (0.69 ± 0.07 h/wk), fast food (times/wk, 0.24 ± 0.05), and breakfast (0.35 ± 0.04 d/wk).


  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgement
  8. References

The Add Health sibling subsample is well-suited for examining the impact of shared household environments on obesity and related behaviors during a lifecycle stage of high-obesity incidence. The contrasts across respondents of varying degrees of genetic relatedness living in shared and non-shared household environments over time provide an important opportunity to isolate the effects of these new and different environmental exposures from the effects of familial, household, and physical environments during childhood and adolescence that persist into young adulthood.

Among MZ and same-sex DZ twin pairs, correlations of behavior patterns of twins living apart were different from those of twins who continued sharing household environments. We hypothesized that such differences reflect the influence of external characteristics on behavior (e.g., physical environment and resource accessibility) (30, 31). For example, compared with twins living together, MZ twins living apart had greater discordance in BMI gain and young adult BMI, and DZ twins had greater longitudinal change in MVPA and young adult fast food intake. In contrast to our hypothesis, however, MZ twins living apart were more highly correlated in longitudinal and young adult sedentary behavior, and DZ twins were more highly correlated in likelihood of achieving ≥5 bouts of MVPA per week. We could not determine the extent to which twins’ non-shared household environments differed in adulthood. Future work should investigate how specific differences in non-shared environments affect these inconsistencies in divergent health patterns.

Overall, our findings highlight important genetic and environmental influences on obesity and related health behaviors. Genetic influences are robust during the adolescent and young adult periods. These results indicate that additive genetic effects account for approximately one-third to one-half of variation in overweight/obesity during this time, as well as 12% to 60% of variation in obesity-related behaviors.

The influence of childhood environment is also important, persisting into young adulthood, even after households are no longer shared. Our results show that the household/physical environment during childhood/adolescence accounted for ∼10% of variation in adolescent fast food intake and sedentary behaviors and ∼50% of adolescent overweight. Interestingly, household effects on MVPA were substantially greater in young adulthood (accounting for ∼50% of variation) compared with adolescence (when household effects accounted for <10% of variation). Our results underscore the importance of household and physical environments as determinants of behavior, leaving a lasting impression on youth that may affect future decision-making and the adoption of healthy behaviors in new environments later in life.

Moreover, our findings indicate that environmental effects may vary, becoming more influential in determining phenotypic extremes (as evidenced by the increasing household component accounting for variation in BMI-for-age z scores, risk for overweight, and overweight, respectively). These findings suggest that in conducting genome-wide linkage studies to identify obesity-related quantitative trait loci, the practice of specifically targeting pedigrees with phenotypic extremes in childhood (32) may not be helpful because the effects of environmental influences seem to be larger in individuals with phenotypic extremes during adolescence.

Although our findings show a substantial influence of young adult household environments on frequency of fast food intake (accounting for ∼12% of variation), other significant effects of households were not detected. At this stage of the life course, however, individuals may have been exposed to their new environments for relatively short durations. The environments of young adults may have not been stable for a long enough time period to substantially influence these behaviors. Therefore, detecting environmental effects on some behaviors may be possible only later in adulthood. Future research is needed to assess individuals in mid-adulthood, where environmental effects may be more pronounced.

Interpretation of our findings requires additional caveats. Firstly, our diet measures are broad indicators of dietary patterning and do not provide quantification of energy balance, which is likely under both genetic and environmental control. Furthermore, our findings show a noteworthy impact of unmeasured/unknown (residual) effects on outcomes of interest, especially behaviors that are difficult to measure with precision (e.g., activity and dietary characteristics, in comparison with BMI). In particular, between Waves II and III, genetic and household effects generally decreased, whereas residual effects increased. These residual effects may represent several different factors including differential responses of individuals to similar exposures (e.g., household/neighborhood environmental exposures) and/or measurement error. More importantly, residual variation may jointly affect numerous weight-related phenotypes, as illustrated by our bivariate analyses, and, thus, may be important to explore.

In addition, our estimates of genetic effects may be marginally inflated. For example, twins may share additional household, parental, or other influences that other siblings living in the same household do not share, thus violating any equal-environment assumptions (11). In our analyses, these effects would be included as additive genetic effects. Our study design is limited by not sampling related individuals living in different adolescent households, thus reducing our ability to precisely distinguish genetic and adolescent household effects and potentially influencing effect estimate stability. Thus, we suggest that future work build on twin and extended pedigree models to include detailed measures characterizing specific attributes of shared and non-shared environments (among related and unrelated individuals) to better understand determinants of health behavior, tease apart residual effects, and describe factors that may vary between siblings in a family.

Despite these caveats, our findings provide important insight into obesity etiology and, to our knowledge, are the first to use these methods in a large longitudinal sibling sample to quantify environmental and genetic influences on BMI and health behavior during this critical life stage. Although a strong genetic component of resting energy expenditure has been previously established by twin and family studies (e.g., (33)), the importance of heritability in PA and sedentary behavior has been less studied. Familial resemblance in physical performance, exercise response, and long-term exercise frequency has been demonstrated in adults (34, 35, 36, 37) and, to a more limited extent, in children (38, 39), although the literature has not been entirely consistent in documenting these similarities (40). In adolescence, we estimated that the variance accounted for by additive genetic effects of PA was 60% and of adolescent weight status, fast food intake, and sedentary behavior was 30% to 35%. Genetic effects on these factors were similar (although somewhat lower) in young adulthood, except for genetic effect on BMI (estimated to be >50%).

The transition from adolescence to adulthood is a critical life course stage, reflecting a phenomenal jump in obesity risk (1). These results illustrate important environmental influences on weight gain, PA, and dietary characteristics during the transition to adulthood, particularly highlighting the substantial role that household and physical environments play in establishing healthful behavior patterns that can persist even after young adults have established new households that are independent of their parents and family. These findings underscore the increasing need for obesity prevention strategies aimed at the household environment in childhood and adolescence.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgement
  8. References

This work was supported by the National Institute of Child Health and Human Development and by the Centers for Disease Control and Prevention (Grants R01-HD39183, R01-HD041375, P30-HD05798, K01-HD044263-01, and CDC-U48/CCU409660). This research uses data from Add Health, a program project designed by J.R. Udry, P.S. Bearman, and K.M. Harris (National Institute of Child Health and Human Development) with cooperative funding from 17 other agencies. Special acknowledgment is due R.R. Rindfuss and B. Entwisle for assistance in the original design. Persons interested in obtaining data files should contact Add Health, Carolina Population Center, 123 W. Franklin Street, Chapel Hill, NC 27516-252 (www.cpc.unc.eduprojectsaddhealthdatacontract).

  • 1

    Nonstandard abbreviations: PA, physical activity; MZ, monozygotic; DZ, dizygotic; MV, moderate-vigorous.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgement
  8. References
  • 1
    Gordon-Larsen, P., Adair, L. S., Nelson, M. C., Popkin, BM. (2004) Five-year obesity incidence in the transition period between adolescence and adulthood. Am J Clin Nutr. 80: 569575.
  • 2
    Gordon-Larsen, P., Nelson, M. C., Popkin, BM. (2004) Longitudinal physical activity and sedentary behavior trends. Am J Prev Med. 27: 277283.
  • 3
    Frank, L. D., Andresen, M. A., Schmid, TL. (2004) Obesity relationships with community design, physical activity and time spent in cars. Am J Prev Med. 27: 8796.
  • 4
    Davidson, K. K., Birch, LL. (2001) Child and parent characteristics as predictors of change in girls’ body mass index. Int J Obes Relat Metab Disord. 25: 18341842.
  • 5
    Gordon-Larsen, P., McMurray, R. G., Popkin, BM. (2000) Determinants of adolescent physical activity and inactivity patterns. Pediatrics 105: 18.
  • 6
    Soubhi, H., Potvin, L., Paradis, G. (2004) Family process and parent's leisure time physical activity. Am J Health Behav. 28: 218230.
  • 7
    Bowman, S. A., Gortmaker, S. L., Ebbeling, C. B., Pereira, M. A., Ludwig, DS. (2004) Effects of fast-food consumption on energy intake and diet quality among children in a national household survey. Pediatrics 113: 112118.
  • 8
    Perusse, L., Rankinen, T., Rauramaa, R., Rivera, M. A., Wolfarth, B., Bouchard, C. (2003) The human gene map for performance and health-related fitness phenotypes. Med Sci Sports Exerc. 35: 12481264.
  • 9
    Comuzzie, A. G., Allison, DB. (1998) The search for human obesity genes. Science 280: 13741377.
  • 10
    Bouchard, C. (1996) Long-term porgramming of body size. Nutr Rev. 54: S8S16.
  • 11
    Maes, H. H., Neale, M. C., Eaves, LJ. (1997) Genetic and environmental factors in relative body weight and human adiposity. Behav Genet. 27: 325351.
  • 12
    Allison, D. B., Kaprio, J., Korkeila, M., Koskenvuo, M., Neale, M. C., Hayakawa, K. (1996) The heritability of body mass index among an international sample of monozygotic twins reared apart. Int J Obes Relat Metab Disord. 20: 501506.
  • 13
    Popkin, B. M., Udry, J.R. (1998) Adolescent obesity increases significantly in second and third generation US immigrants. J Nutr. 128: 701706.
  • 14
    Rowe, D., Jacobson, K., van den Oord, E. (1999) Genetic and environmental influences on vocabulary IQ. Child Development 70: 11511162.
  • 15
    Lench, N., Stanier, P., Williamson, R. (1988) Simple non-invasive method to obtain DNA for gene analysis. Lancet 13561358.
  • 16
    Meulenbelt, I., Droog, S., Trommelen, G. J., Boomsma, D. I., Slagboom, PE. (1995) High-yield noninvasive human genomic DNA isolation method for genetic studies in geographically dispersed families and populations. Am J Hum Genet. 57: 12521254.
  • 17
    Freeman, B., Powell, J., Ball, D., Hill, L., Craig, I., Plowmin, R. (1997) DNA by mail: an inexpensive and noninvasive method for collecting DNA samples from widely dispersed populations. Behav Genet. 27: 251257.
  • 18
    Kuczmarksi, R. J., Ogden, C. L., Grummer-Strawn, L. M., et al (2000) CDC Growth Charts: US: Advance Data from Vital and Health Statistics, No. 314 National Center for Health Statistics Hyattsville, MD.
  • 19
    Himes, J. H., Dietz, WH. (1994) Guidelines for overweight in adolescent preventive services: Expert Committee on Clinical Guidelines for Overweight in Adolescent Preventive Services. Am J Clin Nutr. 59: 307316.
  • 20
    WHO Consultation on Obesity (2000) Obesity: Preventing and Managing the Global Epidemic: Technical Report Series 894 World Health Organization Geneva, Switzerland.
  • 21
    Sallis, J. F., Buono, M. J., Roby, J. J., Micale, F. G., Nelson, JA. (1993) Seven-day recall and other physical activity self-reports in children and adolescents. Med Sci Sport Exerc. 25: 99108.
  • 22
    Gordon-Larsen, P., McMurray, R. G., Popkin, BM. (1999) Adolescent physical activity and inactivity vary by ethnicity: The National Longitudinal Study of Adolescent Health. J Pediatr. 135: 301306.
  • 23
    Andersen, R. E., Crespo, C. H., Bartlett, S. J., Cheskin, L. J., Pratt, M. (1998) Relationship of physical activity and television watching with body weight and level of fatness among children: results from the Third National Health and Nutrition Examination Survey. JAMA 279: 938942.
  • 24
    Pate, R. R., Heath, G. W., Dowda, M., Trost, SG. (1996) Associations between physical activity and other health behaviors in a representative sample of US adolescents. Am J Publ Health. 86: 15771581.
  • 25
    Ainsworth, B., Haskell, W. L., Leon, AS. (1993) Compendium of physical activities. Med Sci Sport Exerc. 25: 7180.
  • 26
    Pate, R. R., Pratt, M., Blair, S. N., et al (1995) Physical activity and public health: a recommendation from the Centers for Disease Control and Prevention and the American College of Sports Medicine. JAMA 273: 402407.
  • 27
    Blalock, HM. (1972) Social Statistics McGraw-Hill New York, NY.
  • 28
    Almasy, L., Blangero, J. (1998) Multipoint quantitative-trait linkage analysis in general pedigrees. Am J Hum Genet. 62: 11981211.
  • 29
    Lange, K. L., Little, R. J. A., Taylor, JMG. (1989) Robust statistical modeling using the t distribution. J Am Stat Assoc. 85: 881896.
  • 30
    Gordon-Larsen, P., Nelson, M. C., Popkin, BM. (2006) Inequality in the built environment underlies key health disparities. Pediatrics 117: 417424.
  • 31
    Morland, K., Wing, S., Diez Roux, A. (2002) The contextual effect of the local food environment on residents’ diets: the atherosclerosis risk in communities study. Am J Public Health 92: 17611767.
  • 32
    Bell, C. G., Walley, A. J., Froguel, P. (2005) The genetics of human obesity. Nat Rev Genet. 6: 221234.
  • 33
    Goran, MI. (1997) Genetic influences on human energy expenditure and substrate utilization. Behav Genet. 27: 389399.
  • 34
    Perusse, L., Bouchard, C. (2000) Gene-diet interactions in obesity. Am J Clin Nutr. 72: 1285190S.
  • 35
    Rice, T., An, P., Gagnon, J., et al (2001) Heritability of HR and BP response to exercise training in the HERITAGE Family Study. Med Sci Sports Exerc. 34: 972979.
  • 36
    Maia, J. A., Thomis, M., Beunen, G. (2002) Genetic factors in physical activity levels. Am J Prev Med. 23: (Suppl 2), 8791.
  • 37
    Simonen, R. L., Perusse, L., Rankinen, T., Rice, T., Rao, D. C., Bouchard, C. (2002) Familial aggregation of physical activity levels in the Quebec Family Study. Med Sci Sports Exerc. 34: 11371142.
  • 38
    Perusse, L., Tremblay, A., Leblanc, C., Bouchard, C. (1989) Genetic and environmental influences on level of habitual physical activity and exercise participation. Am J Epidemiol. 129: 10121022.
  • 39
    Bouchard, C., Lortie, G., Simoneau, J. A., Leblanc, C., Theriault, G., Tremblay, A. (1984) Submaximal power output in adopted and biological siblings. Ann Hum Biology 11: 303309.
  • 40
    Poehlman, E. T., Tremblay, A., Marcotte, M., Perusse, L., Theriault, G., Bouchard, C. (1987) Heredity and changes in body composition and adipose tissue metabolism after short-term exercise-training. Eur J Appl Physiol Occup Physiol. 56: 398402.