Introduction. TL Nelson 1 *, GP Vogler 1, NL Pedersen 2,3 and TP Miles 4

International Journal of Obesity (1999) 23, 449±455 ß 1999 Stockton Press All rights reserved 0307±0565/99 $12.00 http://www.stockton-press.co.uk/ijo Genetic and environmental in uences on waistto-hip ratio and waist circumference in an older Swedish twin population TL Nelson 1 *, GP Vogler 1, NL Pedersen 2,3 and TP Miles 4 1 Department of Biobehavioral Health and Center for Developmental and Health Genetics, The Pennsylvania State University, 101 Amy Gardner House, University Park, PA 16802, USA; 2 Division of Genetic Epidemiology, Institute of Environmental Medicine, Karolinska Institute, Box 210, S 171 77 Stockholm, Sweden; 3 Department of Psychology, University of Southern California, Los Angeles, CA, USA and 4 Department of Family Practice, Division of Geriatrics, University of Texas Health Sciences Center-San Antonio, 7703 Floyd Curl Drive, San Antonio, TX 78284, USA OBJECTIVE: To investigate the genetic and environmental in uences on waist-to-hip ratio (WHR) and waist circumference (WC) measurements in males and females. DESIGN: Measurements taken from 1989 ± 1991 as part of The Swedish Adoption=Twin Study of Aging (SATSA) were used for analysis. The SATSA sample contains both twins reared together as well as twins reared apart. SUBJECTS: 322 pairs of twins (50 identical, 82 fraternal male pairs and 67 identical, 123 fraternal female pairs); age range: 45 ± 85 y (average age, 65 y). MEASUREMENTS: Waist-to-hip ratio (WHR), waist circumference (WC) and body mass index (BMI). RESULTS: In males, additive genetic effects were found to account for 28% of the variance in WHR and 46% of the variance in WC. In females, additive genetic effects were found to account for 48% of the variance in WHR and 66% of the variance in WC. The remaining variance in males was attributed to unique environmental effects (WHR, 72%; WC, 54%) and in females the remaining variance was attributed to unique environmental effects (WHR, 46%; WC, 34%) and age (WHR, 6%). When BMI was added into these models it accounted for a portion of the genetic and environmental variance in WHR, and over half of the genetic and environmental variance in WC. CONCLUSION: There are both genetic and environmental in uences on WHR and WC, independent of BMI in both males and females, and the differences between the sexes are signi cantly different. Keywords: genetic; environmental; waist-to-hip ratio; waist circumference Introduction The adverse metabolic consequences of adipose tissue located in the abdominal region has been documented extensively. Regional body fat distribution has been associated with non insulin-dependent diabetes mellitus (NIDDM), hypercholesterolaemia, hyperinsulinaemia, insulin resistance, hypertriglyceridaemia, atherosclerosis and hypertension. 1±5 There are several ways to measure body fat distribution including anthropometric measures, computerized tomography (CT), magnetic resonance imaging (MRI) or dual energy X-ray absorptiometry (DEXA). CT scans and MRIs and DEXA are the ideal measures of central abdominal fat, however, their use in large-scale studies is limited due to the expense of these procedures. Anthropometric measures are most commonly used in such large-scale epidemiological studies, although they only provide an indirect estimate of abdominal body fat. *Correspondence: Tracy L. Nelson, Department of Epidemiology, Cardiovascular Disease Program, 137 E. Franklin St, NationsBank Plaza, Suite 306, The University of North Carolina at Chapel Hill, Chapel Hill, NC 27514, USA. E-mail: tnelson@sph.unc.edu Received 21 April 1998; revised 12 November 1998; accepted 2 December 1998 The waist-to-hip ratio (WHR) has been the most commonly used anthropometric measure, however, recent work suggests that using the waist circumference (WC) is more closely associated with measures of abdominal visceral fat, that is, the fat most associated with the metabolic problems listed above. 6±8 There are known to be both genetic and environmental in uences on regional body fat distribution, however, it is still unclear how these in uences may vary among different age groups and in each gender. It is important to obtain estimates for age effects and gender effects, because regional body fat distribution is known to increase with age and average values tend to be quite different for males and females. 9,10 Genetic in uences have been estimated in several studies for WHR. In a family study of Mexican Americans, genetic in uences were found to account for 17.3% of the variance in WHR. 11 Other estimates have found WHR to be 28 ± 50% transmissible; such estimates include both genetic and cultural factors (that is, shared environment). 12,13 There are three twin studies we are aware of that estimated WHR and=or WC in adults. Selby et al, 14 using The National Heart, Lung and Blood Institute's Twin Study (NHLBI) found heritability estimates for WHR (h 2 ˆ 0.31, P ˆ 0.07) and WC (h 2 ˆ 0.46, P ˆ 0.02) after controlling for

450 body mass index (BMI) Cardon et al 15 also used the NHLBI Twin Study data and found a heritability of 46% (12% of this variance was shared with BMI) for WHR. They used maximum likelihood methods which provide more power than the methods used by Selby et al. 14 Rose et al 16 using the Kaiser Permanente Twin Study found a heritability of 61% for WHR and 76% for WC in females, after controlling for BMI and age. This is the rst analysis we are aware of that uses a twin=adoption design to estimate the heritability of WHR and WC. Family studies, as with twin studies, may overestimate genetic effects due to confounding genetic and environmental factors, since relatives living together share environments and genes. Twin studies may overestimate genetic effects if monozygotic (MZ) twins share more environmental in uences than dizygotic (DZ) twins. A design that reduces this kind of bias is one that combines twin studies with adoption studies. The Swedish Adoption=Twin Study of Aging (SATSA) uses this design by comparing MZ and DZ twins reared together (MZT and DZT, respectively) and twins reared apart (MZA and DZA, respectively). The purpose of this analysis was to determine the genetic and environmental components of variance in body fat distribution, measured by WHR and WC, in Swedish men and Swedish women with an average age of 65 y. Methods Data for this study came from the Swedish Adoption=Twin Study of Aging (SATSA). The SATSA sample was identi ed through the Swedish Twin Registry, which includes almost 25 000 pairs of same-sexed twins born in Sweden during 1886 ± 1958. 17 The SATSA subregistry was formed in 1984 by contacting twin pairs identi ed in the Swedish Twin Registry as having been reared apart, along with a matched sample of twins reared together. Those pairs reared apart were identi ed based on their responses to the item: `How long did you and your twin partner live together in the same home?' All pairs in which one or both members indicated that they were separated from each other prior to their 11th birthday were included for further study. The identi- cation and characterization of the SATSA sample has been described in detail elsewhere. 18 Measurements of WHR and WC were obtained from a subset of 322 pairs of twins which includes pairs where both twins may not have information on WHR and WC, who underwent physical examinations between 1989 ± 1991. The average age of twins used in this analysis was 65 y (range 45 ± 85 y) with 60% female and 40% male. Measures WC measurements were obtained at the circumference around the smallest part of the waist, and hip measurements as the circumference around the widest point between hips and buttocks. WHR was then determined by dividing the waist measurement by the hip measurement. Age and gender were determined by self-reported information during screening in 1984. BMI was calculated as weight (kg) divided by height squared (m 2 ). Height (m) and weight (kg) were obtained from subjects dressed in lightweight clothes with their shoes removed. Statistical analysis Analyses included descriptive statistics and genetic analyses. Descriptive analyses were performed using SAS, 19 and included means and standard deviations as well as correlations of WHR, WC and BMI. Genetic analyses were performed using Mx 20 to evaluate quantitative contributions of genetic (additive and dominant) and environmental (shared and unique) components. The assumptions using this model- tting analysis are that MZ twins share 100% of their additive and dominant genetic effects and DZ twins share 50% of their additive genetic effects and 25% of their dominant genetic effects. Twins reared together share similar rearing environmental effects and those reared apart do not share rearing environmental effects. This sample allowed us to partition out dominant, as well as additive, genetic effects, because the SATSA sample contains twins reared together in addition to twins reared apart (see Figure 1). In traditional twin studies, which only contain twins reared together, shared rearing environmental and dominant effects confound each other and therefore cannot be estimated separately. In large multivariate samples, there is often missing data. One twin may not have data for the variable being studied. To avoid having to discard all data from such pairs for analysis, we used Mx, 20 a model- tting program that allows missing data to be considered in the analysis. Since Mx uses raw data instead of variance-covariance matrices, the program does not give an actual t statistic (that is, w 2 ) for the overall model but gives instead a maximum likelihood statistic (minus two times the log-likelihood), based on the multivariate normal function. Relative t of nested models can be evaluated by rst determining the maximum likelihood statistic for a general model and then comparing to a more constrained model. The difference between minus twice the log-likelihood of each model is distributed as a w 2 with degrees of freedom being the difference in the number of parameters estimated in the two different models. So

Table 1 Mean values and standard deviations (s.d.) for waistto-hip ratio (WHR), waist circumference (WC) and body mass index (BMI) in 322 pairs of twins by gender, zygosity and rearing status 451 Males Females Variable Mean s.d. Mean s.d. Figure 1 Univariate genetic model. Path diagram shows genetic and environmental effects on Trait 1. MZT ˆ monozygotic twins reared together; MZA ˆ monozygotic twins reared apart; DZT ˆ dizygotic twins reared together; DZA ˆ dizygotic twins reared apart; G ˆ genetic factor; D ˆ dominant genetic factor; ES ˆ shared environmental factor; E ˆ unique environmental factor; Trait 1 indicates trait in twin 1 and trait in twin 2; and Age indicates age in years for twin 1 and twin 2. for example, if shared environmental effects were set to zero and compared to the general model (difference between minus twice the log-likelihoods), a statistically signi cant w 2 would mean shared environmental effects were a signi cant component of the variance for the variable under consideration. The relative importance of genetic and environmental in uences on WHR and WC can be calculated by squaring and summing the parameter estimates of components of variance for each measure and dividing each squared parameter estimate by the sum of all the squared estimates of components of variance. Since both age and gender are known to affect WHR and WC, the analyses considered age as a covariate in the model and gender differences were assessed. Since BMI was highly correlated with both WHR and WC (males BMI ± WHR ˆ 0.53 P < 0.0001, BMI ± WC ˆ 0.84 P < 0.0001; females BMI ± WHR ˆ 0.40 P < 0.0001, BMI ± WC ˆ 0.87 P < 0.0001), we also ran bivariate analyses with BMI in the model, to determine genetic and environmental variance in WHR and WC, independent of BMI. WHR MZT 0.93 (27) 0.05 0.81 (43) 0.06 MZA 0.95 (23) 0.06 0.81 (24) 0.05 DZT 0.91 (43) 0.05 0.81 (49) 0.06 DZA 0.93 (39) 0.05 0.81 (74) 0.06 WC MZT 96.28 (27) 9.80 82.20 (43) 10.22 MZA 98.21 (23) 8.54 83.05 (24) 8.73 DZT 92.57 (43) 8.23 82.79 (49) 9.00 DZA 97.03 (39) 8.31 84.46 (74) 12.53 BMI MZT 26.13 (27) 3.38 24.96 (43) 3.72 MZA 26.07 (23) 3.13 25.05 (24) 3.80 DZT 24.61 (43) 2.64 25.41 (49) 4.00 DZA 26.16 (39) 3.10 26.20 (74) 5.34 The number of pairs are in parenthesis. Note: these are not all full pairs (that is, variables are included where we only have information for one twin). MZT ˆ monozygotic twins reared together; MZA ˆ monozygotic twins reared apart; DZT ˆ dizygotic twins reared together; DZA ˆ dizygotic twins reared apart. Table 2 Intra-pair correlations and 95% con dence intervals (CI) for waist-to-hip ratio (WHR), waist circumference (WC) and body mass index (BMI) in males and females by zygosity and rearing status Variable Males N 95% CI Females N 95% CI WHR MZT 0.48 27 (0.12, 0.73) 0.65 30 (0.38, 0.82) MZA 0.38 19 ( 7 0.09, 0.71) 7 0.002 19 ( 7 0.45, 0.45) DZT 7 0.12 36 ( 7 0.43, 0.22) 0.30 37 ( 7 0.03, 0.57) DZA 7 0.07 32 ( 7 0.41, 0.28) 0.29 61 0.04, 0.51) WC MZT 0.52 27 (0.17, 0.75) 0.64 30 (0.35, 0.81) MZA 0.64 19 (0.26, 0.84) 0.48 19 (0.03, 0.77) DZT 0.06 36 ( 7 0.28, 0.38) 0.43 37 (0.12, 0.66) DZA 0.08 32 ( 7 0.27, 0.42) 0.32 61 (0.07, 0.68) BMI MZT 0.67 27 (0.39, 0.84) 0.63 30 (0.34, 0.80) MZA 0.55 19 (0.13, 0.80) 0.65 19 (0.28, 0.86) DZT 0.33 36 (0.00, 0.59) 0.50 37 (0.21, 0.70) DZA 0.22 32 ( 7 0.14, 0.53) 0.38 61 (0.14, 0.57) Note: these are all full pairs (pairs where one twin had missing data were deleted to obtain the correlations). MZT ˆ monozygotic twins reared together; MZA ˆ monozygotic twins reared apart; DZT ˆ dizygotic twins reared together; DZA ˆ dizygotic twins reared apart. Results Sample characteristics Female subjects were aged, on average, 67 9 y and male subjects were aged, on average, 63 8 y. Table 1 lists mean levels of WHR, WC and BMI by gender and rearing status. The mean values are about average for this age range in the Swedish population. Table 2 shows the correlations by gender, zygosity and rearing status for WHR, WC and BMI. In general, intra-pair correlations for MZ twins were higher than those for DZ twins, indicating genetic in uences. There were differences in correlations between males and females indicating possible gender differences in heritability for abdominal fatness. The negative correlations for WHR in the male DZA and DZT and female MZA twins might suggest the twins' WHR scores are in opposite directions, however these correlations were not signi cantly different from zero. In traditional genetic analysis, such correlations would limit our ability to assess genetic in uences, because all the information contained in all groups, regarding genetic in uences, is not used

452 simultaneously. However, we used model- tting analyses, which permits analysis of groups of twins simultaneously. This method is more powerful at detecting genetic effects because it uses all of the information in all of the groups in a single comprehensive analysis. Quantitative genetic analysis To assess whether males and females should be considered separately in this genetic analysis, a constrained model was tested, where the parameter estimates were set equal for males and females (see Table 3 and Table 4). These models were signi cantly worse than the full models (WHR: w 2 ˆ 18.05, df ˆ 6, P < 0.01; WC: w 2 ˆ 27.18, df ˆ 6, P < 0.001) indicating males and females had signi cantly different Table 3 Test of models for genetic and environmental in uences on waist-to-hip ratio (WHR) Difference 7 2 log- WHR likelihood df w 2 df P-value AIC 1. Full model 621.978 895 2. Constrained model (males ˆ females) 640.027 901 18.05 6 0.01 6.05 3. Constrained model (no D) 624.461 897 2.48 2 NS 7 1.52 4. Constrained model (no A) 622.558 897 0.58 2 NS 7 3.42 5. Constrained model (no Es) 622.238 897 0.26 2 NS 7 3.74 6. Constrained model (no D, A or Es) 645.068 901 23.09 6 0.001 11.06 7. Constrained model (no Es or D) 624.744 899 2.77 4 NS 7 5.23 8. Constrained model (no A or D) 634.634 899 12.66 4 0.05 4.66 AIC ˆ Akaike's information criterion; NS ˆ not statistically signi cant; A ˆ additive genetic effects; D ˆ dominant genetic effects; Es ˆ shared environmental effects. Table 4 Test of models for genetic and environmental in uences on waist circumference (WC) Difference 7 2 log- WC likelihood df w 2 df P-value AIC 1. Full model 6570.301 893 2 Constrained model (males ˆ females) 6597.476 899 27.18 6 0.001 15.18 3. Constrained model (no D) 6572.270 895 1.97 2 NS 7 2.03 4. Constrained model (no A) 6573.292 895 2.99 2 NS 7 1.01 5. Constrained model (no Es) 6571.677 895 1.38 2 NS 7 2.62 6. Constrained model (no D, A or Es) 6618.562 899 48.26 6 0.001 36.26 7. Constrained model (no Es or D) 6573.712 897 3.41 4 NS 7 4.59 8. Constrained model (no A or D) 6593.754 897 23.45 4 0.001 15.45 AIC ˆ Akaike's information criterion; NS ˆ not statistically signi cant; A ˆ additive genetic effects; D ˆ dominant genetic effects; Es ˆ shared environmental effects. parameter estimates for measures of WHR and WC. A reduced model, where dominant genetic loadings were set to zero in both males and females, had no signi cant loss of t compared to the full model, nor did setting additive genetic effects to zero or shared environmental effects to zero. However, setting dominant and additive genetic effects and shared rearing environmental effects to zero, in males and females, resulted in a signi cantly worse t of the model. To further test this model, we set only shared rearing environmental effects and dominant genetic effects to zero, and this did not change the t of the model. However, setting additive genetic effects and dominant genetic effects to zero resulted in a signi cantly worse model indicating the importance of genetic effects on WHR and WC. To determine which model was the most parsimonious, we computed the Akaike's Information Criterion (AIC) (w 2-2df). The model with the lowest value of this index ts best, according to the AIC. As can be seen in Table 3 and Table 4, model 7 was the most parsimonious (AIC ˆ 7 5.23 and 7 4.59). Model 7 was also the most parsimonious for univariate genetic analysis of BMI (not presented). Table 5 presents the percentage of genetic and environmental in uences for WHR, WC and BMI. In males, additive genetic effects were found to account for 28% of the variance in WHR and 46% of the variance in WC. In females, additive genetic effects were found to account for 48% of the variance in WHR and 66% of the variance in WC. The remaining variance in males was attributed to unique environmental effects (WHR, 72%; WC, 54%) and in females the remaining variance was attributed to unique environment (WHR, 46%; WC, 34%) and age (WHR, 6%). When BMI was added into the WHR and WC models (Figures 2 and 3), it accounted for a portion of the genetic and environmental variance in WHR and over half of the genetic and environmental variance in WC. To present the nature of the genetic and environmental covariances in another way, in Table 6, we have reported phenotypic correlations for WHR and BMI, as well as WC and BMI, in both males and females. We then broke down the phenotypic correlations into a component due to genetic effects and a component due to environmental effects. See Appendix for a description of the phenotypic correlation calculations. Table 5 Percent genetic and environmental in uences for waist-to-hip ratio (WHR), waist circumference (WC) and body mass index (BMI) Males Females Genetic Environmental Age Genetic Environmental Age WHR 28% 72% - 48% 46% 6% WC 46% 54% - 66% 34% - BMI 57% 43% - 72% 28% -

Table 6 Genetic (G) and environmental (E) components of phenotypic correlations for waist-to-hip ratio (WHR) and body mass index (BMI) as well as for waist circumference (WC) and BMI in males and females Males BMI Females BMI WHR 0.49(G ˆ 0.13; E ˆ 0.36) 0.35(G ˆ 0.20; E ˆ 0.15) WC 0.82(G ˆ 0.41; E ˆ 0.41) 0.86(G ˆ 0.60; E ˆ 0.26) was unique to WHR and 8% was contributed from genetic effects common to BMI, while 14% of the genetic in uence was unique to WC and 52% was contributed from BMI. These results also suggest there are environmental effects unique to WHR, WC and BMI, and that both WHR and BMI, as well as WC and BMI, also have environmental effects in common. 453 Discussion Figure 2 Multivariate genetic model for waist-to-hip ratio (WHR) and body mass index (BMI) in males and females. Path diagram shows genetic (G) and environmental (E) effects on WHR and BMI. G1 is genetic factor 1 which loads on BMI and WHR; G2 is genetic factor 2 which only loads on WHR. E1 is unique environmental factor 1 which loads on BMI and WHR and E2 is unique environmental factor 2 which only loads on WHR. Age indicates age in years for twins 1 and twin 2. Figure 3 Multivariate genetic model for waist circumference (WC) and body mass index (BMI) in males and females. Path diagram shows genetic (G) and environmental (E) effects on WC and BMI. G1 is genetic factor 1 which loads on BMI and WHR; G2 is genetic factor 2 which only loads on WC. E1 is unique environmental factor 1 which loads on BMI and WC and E2 is unique environmental factor 2 which only loads on WC. Overall, there are genetic effects unique to BMI, WHR and WC, and there are also genetic effects in common to BMI and WHR, as well as BMI and WC. In males, 25% of the genetic in uence was unique to WHR and 3% was contributed from genetic effects in common with BMI, while 17% of the genetic in uence was unique to WC and 29% contributed from BMI. In females, 40% of the genetic in uence The purpose of this study, was to estimate the effects of genetic and environmental in uences on regional body fat distribution, as measured by WHR and WC. We found an additive genetic in uence on WHR and WC, in both males and females. This genetic effect was signi cantly different for males and females, and age was a covariate for females. The heritability estimates of WHR and WC are hard to compare to those found in other studies, because of the different methods used. For example, some of the studies used families which in most cases meant they could not separate out genetic effects from cultural or shared environmental effects. The Canadian Fitness Survey showed WHR to be about 28% transmittable (this estimate included genetic and cultural effects). 12 Data from the San Antonio Family Heart Study showed WHR to be 17.3% heritable, but this was with men and women combined, 11 and among women in the Iowa Women's Health Study 13 it was found that the WHR was 40 ± 50% transmittable, but once again they could not separate genes from the shared environment. The NHLBI study by Selby et al 14 suggested a nonsigni cant heritability of 0.31 (P ˆ 0.07) for WHR, but a signi cant heritability for WC (0.46, P ˆ 0.02), however they used the classical estimates based on the intraclass correlations, which have low power and do not use all the information simultaneously. Rose et al 16 found among female twins, in the Kaiser Permanente Twin Study, that WHR was 61% heritable while WC was 76% heritable, after adjusting for BMI and age. The study by Cardon et al 15 is the most comparable to our data, because they used quantitative genetic analysis, however, they only used males. Their results showed that genetic effects explained 46% of the variance in WHR, while ours showed 28% of the variance was explained by genetic effects in males. They also looked at what percentage of this variance was shared with BMI and found that 12% of the 46% genetic variance in WHR was explained by BMI, while our study found that 3% of the 28% genetic variance was explained by BMI. The differences in these ndings may be due to the potential for shared environmental effects to confound genetic effects and thus over-estimate heritabilities when quantitative analyses are made with only twins reared together. A twin study by Carey et al 21 used DEXA scans to measure central abdominal fat and

454 they have reported a heritability estimate of 70% among females, after controlling for age and total body fat. DEXA scans provide a direct measure of abdominal fat and this estimate may be the most accurate of the above. Based on these studies, it is clear there are genetic and environmental in uences on regional fat distribution, independent of overall body fat. The present study gives more insight into the genetic effects on measures of WHR and WC, as model- tting analyses permit the analysis of groups of twins simultaneously. We were able to separate shared environmental effects from genetic effects and estimates were obtained for males and females separately. This study also provides estimates of effects in common with BMI. After adding BMI to the model we found some genetic effects in common, indicating that some of the same genetic effects that are in uencing BMI are also in uencing WHR and WC. Of the genetic variance in WHR, < 25% is in common with BMI in males and females, and over half of the genetic variance in WC is in common with BMI in males and females. The differences may be due to the nature of the measures, as we found WC to be more highly correlated to BMI than WHR (males, BMI-WC ˆ 0.84, P < 0.0001, BMI-WHR ˆ 0.53, P < 0.0001; females, BMI- WC ˆ 0.87 P < 0.0001, BMI-WHR ˆ 0.40, P < 0.0001). Other research has found some effects in common for abdominal fatness and overall fatness, as well as a single genetic effect that may be unique to abdominal fatness. 15,22,23 There also appears to be a percentage of environmental effects in common for BMI and WHR, and over half of the environmental effects are in common between BMI and WC in males and females. Such common environmental in uences may include overeating, lack of physical activity and=or alcohol consumption, as all of these have been found to be associated with indicators of abdominal fat and 24 ± 27 BMI. Conclusion Based on these ndings, it is important to look at genetic and environmental effects of regional body fat distribution, in each gender separately, as well as to look speci cally at age effects. Males tend to have more abdominal fat than women. However, as women become postmenopausal, it is thought that these differences become less pronounced. We did nd age effects accounted for a small portion of the variance in WHR for females, however, we did not have the power to observe these age effects by cohort (that is, 50 ± 60 y; 60 ± 70 y etc.) nor did we have measures of menopausal status. Future studies may wish to look at whether the length of time women are postmenopausal may be used as an indirect way of assessing how much visceral or abdominal fat they may have. Age is also known to be associated with increases in abdominal fat, as people age, body composition changes, more fat becomes deposited in the abdominal fat deposits and less on the periphery. More work is needed to see if genetic effects are different among different cohorts (that is, 40 ± 50 y 50 ± 60 y, 70 ± 80 y and > 80 y), as well as in each gender among these varying cohorts. If there are different genetic effects at different ages, there may also be different environmental in uences acting through the adulthood and into old age, possibly through gene-environment interactions. Bouchard et al 23 has found a major gene, accounting for 51% of the variance in abdominal fat, as measured by CT scans. Once these genes are localized, genotyping can occur, and these potential genetic and environmental interactions can be studied. Acknowledgements This project has been supported by grants from the National Institute on Aging (AG-04563, AG-10175), the National Heart, Lung and Blood Institute (HL- 55976), the MacArthur Foundation Research Network on Successful Aging and the Swedish Council for Social Research. T. Nelson has been supported by the National Institutes on Aging (AG-10430) and National Institute of Child Health and Human Development (HD-07454). References 1 Larsson B, Svardsudd K, Welin L, Wilhelmsen L, BjoÈrntorp P, Tibblin G. Abdominal adipose tissue distribution, obesity and risk of CVD and death: 13 year follow-up of participants in the study of men born in 1913. Br Med J 1984; 288: 1401 ± 1404. 2 BjoÈrntorp P. Regional fat distribution-implications for type II diabetes. Int J Obes 1992; 16 (Suppl. 4): S19 ± S27. 3 Kissebah A, Peiris A, Evans D. Mechanisms associating body fat distribution to glucose intolerance and diabetes mellitus: Window with a view. Acta Med Scand 1988; 723 (Suppl): S79 ± S89. 4 Kissebah A, Krawkower G. Regional adiposity and morbidity. Physiol Rev 1994; 74: 761 ± 811. 5 DespreÂs J, Lemieux S, Lamarche B, Prud' homme D, Moorjani S, Brun LD, Gagne C, Lupien PJ. The insulin resistancedyslipidemic syndrome: contribution of visceral obesity and therapeutic implications. Int J Obes 1995; 19 (Suppl 1): S76 ± S86. 6 Seidell J, Cigolini M, Charzewska J, Ellsinger BM, Deslypere J, Cruz A. Fat distribution in European men: A comparison of anthropometric measurements in relation to cardiovascular risk factors. Int J Obes 1992; 16: 17 ± 22. 7 Pouliot MC, DespreÂs JP, Lemieux S, Moorjani S, Bouchard C, Tremblay A, Nadeau A, Lupien P. Waist circumference and abdominal sagittal diameter: Best simple anthropometric indexes of abdominal visceral adipose tissue accumulation and related cardiovascular risk in men and women. Am J Cardiol 1994; 73: 460 ± 468. 8 van der Kooy K, Seidell J. Techniques for the measurement of visceral fat: A practical guide. Int J Obes 1993; 17: 187 ± 196. 9 Schwartz R, Shuman W, Bradbury V, Cain K, Fellingham G, Beard J, Kahn S, Stratton J, Cerqueira M, Abrass I. Body fat distribution in healthy young and older men. J Gerontol A Bio Sci Med Sci 1990; 45: M181 ± M185. 10 Duncan BB, Chambless LE, Schmidt MI, Szklo M, Folsom AR, Carpenter MA, Crouse JR. Correlates of body fat distribution: Variation across categories of race, sex, and body mass in the atherosclerosis risk in communities study. Ann Epidemiol 1995; 5: 192 ± 200.

11 Mitchell B, Kammerer M, Mahaney M, Blangero J, Comuzzie A, Atwood L, Haffner S, Stern M, MacCluer J. Genetic analysis of the IRS: Pleiotropic effects of genes in uencing insulin levels on lipoprotein and obesity measures, Arterioscler Thromb Vasc Biol 1996; 16: 281 ± 288. 12 Perusse L, Leblanc C, Bouchard C. Inter-generation transmission of physical tness in the Canadian population. Can J Sport Sci 1988; 13: 8 ± 14. 13 Sellers T, Drinkard C, Rich S., Potter J., Jeffery R., Hong C., Folsom A. Familial aggregation and heritability of WHR in adult women: The Iowa Women's Health study. Int J Obes 1994; 18: 607 ± 613. 14 Selby J, Newman B, Quesenberry C, Fabsitz R, Carmelli D, Meaney J, Slemenda C. Genetic and behavioral in uences on body fat distribution. Int J Obes 1990; 14: 593 ± 602. 15 Cardon L. Carmelli D, Fabsitz R, Reed T. Genetic and environmental correlations between obesity and body fat distribution in adult male twins. Hum Biol 1994; 66(3): 465 ± 479. 16 Rose K, Newman B, Mayer-Davis E, Selby J. Genetic and behavioral determinants of waist-hip ratio and waist circumference in women twins. Obes Res 1998; 6: 383 ± 392. 17 Cederlof R, Lorich U. The Swedish twin registry. In: Nance WE, Allen G, Parisi P (eds). Twin research: part C. Biology and Epidemiology. Alan R Liss: New York, 1978, 189 ± 195. 18 Pedersen NL, McClearn GE, Plomin R, Nesselroade JR, Berg S, de Faire U. The Swedish Adoption Twin Study of Aging: An update. Acta Genet Med Gemellol (Roma) 1991; 40: 7 ± 20. 19 SAS Institute, Inc. SAS=STAT, Version 6. Vol. 1. (4th edn). SAS Institute, Inc: Cary NC, 1989. 20 Neale MC. Mx: Statistical Modeling. (3rd edn). Department of Psychiatry: Box 710 MCV, Richmond 1995. 21 Carey DGP, Nguyen TV, Campbell LV, Chisholm DJ, Kelly P. Genetic in uences on central abdominal fat: A twin study. Int J Obes 1996; 20: 722 ± 726. 22 Perusse L, DespreÂs J, Lemieux S, Rice T, Rao D, Bouchard C. Familial aggregation of abdominal visceral fat level: Results from the Quebec family study. Metabolism 1996; 45: 378 ± 382. 23 Bouchard C, Rice T, Lemieux S, DespreÂs J, Perusse L, Rao D. Major gene for abdominal visceral fat area in the Quebec Family Study. Int J Obes 1996; 20: 420 ± 427. 24 Slattery ML, McDonald A, Bild DE, Caan BJ, Hilner JE, Jacobs DR, Liu K. Associations of body fat and its distribution with dietary intake, physical activity, alcohol and smoking in blacks and whites. Am J Clin Nutr 1992; 55: 943 ± 949. 25 Miller WC, Lindman AK, Wallace J. Niederpruem M. Diet composition, energy intake, and exercise in relation to body fat in men and women. Am J Clin Nutr 1990; 52: 426 ± 430. 26 Seidell J. Environmental in uences on regional fat distribution, Int J Obes 1991; 15: 31 ± 35. 27 Kuczmarski R, Flegal K, Campbell S, Johnson C. Increasing prevalence of overweight among US adults: The national health and nutrition examination surveys, 1960 ± 1991. JAMA 1994; 272: 205 ± 211. 455 Appendix Calculation of phenotypic, genetic and environmental correlations The following is the equation used to calculate the phenotypic correlations, as well as the genetic and environmental components of this phenotypic correlation. r P ˆ h x h y r Gxy e x e y r Exy Where r P ˆ phenotypic correlation between variable x and variable y; h x is the square root of the heritability for variable x, and h y is the square root of the heritability for variable y; r Gxy is the genetic correlation between variables x and y. This is calculated as: p Cov Gx Gy V Gx V Gy which is the genetic covariance of x and y divided by the product of the square root of the genetic variance of x and the square root of the genetic variance of y. The environmental correlation is calculated the same way.