Genetic factors that influence BMI have been well documented over the past decades, indicating heritability estimates in the range of 0.5 to 0.9 (1). However, because these estimates came mostly from twins from Western countries, heritability estimates of BMI in Asian populations are largely unknown. A few existing studies using Chinese twins (2) and Japanese twins (3) indicated the importance of genetic factors on BMI. However, these studies were all on the basis of small samples. To better understand genetic and environmental influences on BMI in Asian populations, a large-scale Asian twin study is necessary.

The aim of this study was to evaluate genetic and environmental influences on BMI in South Korean adolescent twins, with a special emphasis on sex difference. In total, 888 adolescent twin pairs (13 to 19 years of age; Table 1) drawn from the South Korean Twin Registry (4) reported height and weight through a telephone survey (N = 187 pairs) and a mail survey (N = 701 pairs) in 2006. The means and standard deviations for BMI in the telephone and mail surveys were not significantly different (data not shown), suggesting that the method of data collection did not affect the results.

Table 1 - Sample sizes, means, and SDs of age and BMI, twin correlations (r), and their 95% CIs for five groups of twins.

Full table

To our knowledge, the sample for this study contained the largest number of Asian twins ever reported in twin studies of BMI. Twins' zygosity in the South Korean Twin Registry was determined from the twins' parents' responses to a zygosity questionnaire. This questionnaire method to determine zygosity has been widely used in large twin studies and is regarded as >90% accurate for twins in general (5). However, in this study, cases where zygosity could not be classified with certainty were excluded from data analyses.

BMI was calculated as weight in kilograms divided by height in meters squared. The distribution of BMI was somewhat positively skewed (skewness = 0.73, kurtosis = 1.03). The mean and standard deviation for BMI in the total sample were 20.1 and 2.6, respectively.

Maximum likelihood correlations were 0.83 [ 95% confidence interval (CI)¹: 0.79 to 0.86] for monozygotic male (MZM) twins, 0.33 (95% CI: 0.12 to 0.51) for dizygotic male (DZM) twins, 0.88 (95% CI: 0.85 to 0.90) for monozygotic female (MZF) twins, 0.42 (95% CI: 0.22 to 0.58) for dizygotic female (DZF) twins, and 0.31 (95% CI: 0.14 to 0.46) for opposite-sex dizygotic (OSDZ) twins. For both sexes, maximum likelihood monozygotic (MZ) twin correlations were significantly higher than dizygotic (DZ) twin correlations, indicating substantial genetic effects and negligible shared environmental influence on BMI. However, OSDZ twin correlation was not significantly different from correlations for DZM or DZF twins, suggesting little evidence for sex-specific genetic effects.

A general sex-limitation model was applied to the data using the raw data option in Mx (6), a statistical software designed for twin and family data. The difference in ² between the general sex-limitation model and the saturated model was not significant (² = 23.8, for 44 df, p > 0.97), suggesting that the data do not depart significantly from the general sex-limitation model. In the general sex-limitation model, sex and age variables were treated as covariates to control for their main effects. Significant changes in ² occurred when sex or age was removed from the model (² = 18.0 for 1 df, p < 0.01 for sex; ² = 32.9 for 1 df, p < 0.01 for age). Additive and non-additive genetic correlations for OSDZ twins (r_aO and r_dO, respectively) could be fixed at 0.5 and 0.25, respectively, without a significant change in ² (² = 0.0 for 2 df, p > 0.99). These results indicate that the same genes might be operating in men and women. In subsequent reduced models, therefore, r_aO and r_dO were constrained to be 0.5 and 0.25, respectively. Next, additive genetic (A), non-additive genetic (dominance; D), and non-shared environmental (E) parameters were equated across men and women in various ways. These procedures consistently yielded significant changes in ², indicating that the magnitudes of genetic and environmental influences on BMI are not equal across the sexes. Finally, the D parameter was eliminated from the model, while allowing A and E parameters to vary across sexes. Resulting ² change was not significant (² = 3.3 for 4 df, p = 0.51). Under this model, additive genetic and non-shared environmental variance estimates were 82% (95% CI: 72% to 95% ) and 18% (95% CI: 15% to 21% ), respectively, for men and 87% (95% CI: 77% to 99% ) and 13% (95% CI: 11% to 15% ), respectively, for women (Table 2).

Table 2 - Standardized and unstandardized additive and non-additive genetic and non-shared environmental variance estimates and their 95% CIs in the full and best-fitting model.

Full table

Although the heritability estimate was significantly larger in women than in men, the total unstandardized phenotypic, genetic, and non-shared environmental variances were significantly larger in men than in women (² = 39.2 for 9 df, p < 0.00). These results suggest a possibility that there might be genes and environmental factors acting specifically in men only. However, as explained earlier, the OSDZ twin correlation in this sample was not significantly lower than the same-sex DZ twins. These conflicting results need to be resolved in the future with a larger twin sample.

The classic twin design used in this study assumes that there are no intrauterine environmental effects. Given the correlations between birth weight and adult and adolescent BMI reported in previous studies (7), however, the intrauterine environment may be critical for the development of BMI. A recent study (8) has shown that the relationship between birth weight and adult BMI is largely mediated by correlated genetic and non-shared environmental factors. These results suggest that intrauterine environmental effects on BMI, if they exist, are largely determined by genetic factors and support a lack of shared environmental effects on BMI found in this sample.

The measure of BMI used in this study was based on self-reported height and weight. Height for each age group was very similar to the national average height; however, weight was generally lower than the national average weight for each age group (9), suggesting that twins in this sample may have under-reported their weights. This bias may have resulted in slightly lowered estimates of heritability in this study.

To conclude, this study showed that genetic influences on BMI in South Koreans are substantial and predominantly additive rather than non-additive and that the estimates of genetic and environmental factors in BMI are broadly in the range of those reported in previous studies of BMI based on Western twins. Consistent with several studies (10, 11, 12, 13), heritability of BMI was slightly, but significantly, higher among women than men in this sample. This study also confirmed previous reports of negligible shared family environmental factors in BMI. These findings are interesting given the well-documented racial/ethnic differences in gene frequencies (14), dietary practice, health behaviors (15), and obesity-related metabolic syndromes (16). Future studies should investigate how genes for obesity that vary in frequency among populations interact with the environment in different populations and how these interactions lead to a large difference in the phenotypic distribution of BMI among various populations.

Research Methods and Procedures

To study sex difference in genetic and environmental influences on BMI, maximum likelihood correlations were computed for five groups of twins (MZM, DZM, MZF, DZF, and OSDZ) and model-fitting analyses were carried out. Mx (6) was used to conduct both correlational and model-fitting analyses.

Figure 1 depicts a path diagram of the general sex-limitation model used for this study. The total phenotypic variance of BMI is decomposed into three sources: A, D, and E factors. Measurement error is confounded with the E factors. Shared environmental factors were not considered in this model because, consistent with previous studies based on many Western twins, the twin correlations found in this study showed negligible influences of shared environmental factors in BMI. The A factors, the sum of the average effect of all genes that influence a trait, correlate at 1.0 and at 0.5 for MZ and same-sex DZ twins, respectively. The D factors, the genetic factors that do not add up across genes because of allelic interaction, correlate at 1.0 for MZ and at 0.25 for same-sex DZ twins. For OSDZ twins, however, the correlations for A and D factors were allowed to vary between 0 and 0.50 and between 0 and 0.25, respectively, assuming that some of the genes that determine BMI may be different between men and women. Finally, the E factors, environmental factors that are unique to each member of a twin pair and measurement error, do not contribute to the twin similarity and, therefore, are depicted in the path diagram as residual arrows for each twin representing the remaining variance not explained by additive or non-additive genetic factors. A, D, and E parameters were allowed to differ between men and women (A_m A_f, D_m D_f, and E_m E_f), assuming that the magnitudes of additive and non-additive genetic influences and non-shared environmental influences on BMI may vary in men and women. Underlying assumptions for the sex-limitation model and the twin method can be found in Plomin et al. (17) and Neale and Cardon (18).

Figure 1:.

General sex limitation model. The variables A, D, and E refer to additive genetic factors, non-additive genetic factors, and non-shared environmental factors, respectively. The m and f subscripts refer to males and females, respectively. r_a and r_d are additive genetic and non-additive genetic correlations between same sex twins, respectively; r_aO and r_dO are additive genetic and non-additive genetic correlations between opposite sex twins.

Full figure and legend (89K)

First, the fit of the full sex-limitation model in Figure 1 was compared with that of saturated model where variances and means of the first- and the second-born MZ and DZ twins were allowed to vary. Next, the fit of the full sex-limitation model was compared with the fit of a series of reduced models. Three different kinds of constraints were made in the reduced models. First, additive and non-additive genetic correlations for OSDZ twins were fixed to be 0.5 and 0.25, respectively, to examine the presence of the sex-specific genes. Second, the magnitudes of A, D, and E parameters (both standardized and unstandardized variance parameters) were equated across sexes to study the sex difference in the estimates of genetic and environmental influences on BMI. Finally, A or D parameter (both standardized and unstandardized variance parameters) was eliminated from the full model to determine the significance of each parameter. The E parameter was not removed because measurement error was confounded with the E parameter.

The raw data option in Mx (6) calculates twice the negative log-likelihood (–2LL) of the data. Because the difference in –2LL between the two hierarchically related models (i.e., the full vs. reduced model) is ² distributed with degrees of freedom equal to the difference in degrees of freedom between the two models, the ² difference test was used to select the best-fitting, most parsimonious model from alternative models. A significant change in ² between the full and reduced model would indicate that the reduction of the model is not acceptable, whereas a non-significant change in ² would suggest the opposite.

Notes

¹ Nonstandard abbreviations: CI, confidence interval; MZM, monozygotic male twins; DZM, dizygotic male twins; MZF, monozygotic female twins; DZF, dizygotic female twins; OSDZ, opposite-sex dizygotic; MZ, monozygotic; DZ, dizygotic; r_aO, additive genetic correlations for OSDZ twins; r_dO, non-additive genetic correlations for OSDZ twins; A, additive genetic; D, dominance; E, non-shared environmental; 2LL, twice the log-likelihood.

References

1. Maes, H. M., Neale, M. C., Eaves, LJ. (1997) Genetic and environmental factors in relative body weight and human adiposity. Behav Genet. 27: 325–351. | Article | PubMed | ISI | ChemPort |
2. Xu, X., Chen, C., Niu, T., et al (2000) Twin and sib-pair studies in developing countries. In Spector, TD Sneider, H and MacGregor, AJ (eds). Advances in Twin and Sib-Pair Analysis. Oxford University: London, UK. 91–102.
3. 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: 501–506. | PubMed | ChemPort |
4. Hur, Y-M, Shin, J. S., Jeong, H-U, Han, JY. (2006) The South Korean Twin Registry. Twin Res Hum Genet. 10: 838–843.
5. Ooki, S., Yamada, K., Asaka, A. (1992) Zygosity diagnosis of twins by questionnaire for twins' mothers. Acta Genet Med Gemellol (Roma). 42: 17–22.
6. Neale, M., Boker, S. M., Xie, G., et al (2003) Mx: Statistical Modeling. Medical College of Virginia, Department of Psychiatry: Richmond, VA.
7. Phillips, D. I. W., Young, JB. (2000) Birthweight, climate at birth and the risk of obesity in adult life. Int J Obes Relat Metab Disord. 24: 281–287. | Article | PubMed | ChemPort |
8. Whitfield, J. B., Treloar, S. A., Zhu, G., et al (2001) Genetic and non-genetic factors affecting birth-weight and adult body mass index. Twin Res Hum Genet. 4: 365–370.
9. Korean National Statistical Office (Accessed 2005) www.nso.go.kr.
10. Pietilainen, K. H., Kaprio, J., Rissanen, A., et al (1999) Distribution and heritability of BMI in Finnish adolescents aged 16 y and 17 y: a study of 4884 twins and 2509 singletons. Int J Obes Relat Metab Disord. 23: 107–115. | Article | PubMed | ChemPort |
11. Allison, D. B., Heshka, S., Neale, M. C., et al (1994) A genetic analysis of relative weight among 4020 twin pairs, with an emphasis on sex effects. Health Psychol. 13: 362–365. | PubMed |
12. Harris, J. R., Tambs, K., Magnus, P. (1995) Sex-specific effects for body mass index in the New Norwegian Twin Panel. Genet Epidemiol. 12: 251–265. | Article | PubMed | ISI | ChemPort |
13. Herskind, A. M., McGue, M., Sorensen, TIA. (1996) Sex and age specific assessment of genetic and environmental influences on body mass index in twins. Int J Obes Relat Metab Disord. 20: 106–113. | PubMed |
14. Goldenberg, I., Moss, A. J., Ryan, D., McNitt, S., Eberly, S. W., Zareba, W. (2006) Polymorphism in the angiotensinogen gene, hypertension, and ethnic differences in the risk of recurrent coronary events. Hypertension. 48: 693–699.
15. Fernandez, J. R., Allison, DB. (2003) Understanding racial differences in obesity and metabolic syndrome traits. Nutr Rev. 61: 316–319.
16. Ikegami, H., Fujisawa, T., Kawabata, Y., Noso, S., Ogihara, T. (2006) Genetics of type 1 diabetes: similarities and differences between Asian and Caucasian populations. Ann N Y Acad Sci. 1079: 51–59. | Article | PubMed | ISI | ChemPort |
17. Plomin, R., DeFries, J. C., McGuffin, P. (2000) Behavior Genetics: A Primer. Freeman: New York, NY.
18. Neale, M. C., Cardon, LR. (1992) Methodology for Genetic Studies of Twins and Families. Kluwer: London, UK.

Acknowledgments

Financial support was provided, in part, by Korean Research Foundation (KRF2001 041 c00548).