Exercise behavior, cardiorespiratory fitness, and obesity are strongly influenced by genetic factors. By studying young adult twins, we examined to what extent these interrelated traits have shared genetic and environmental etiologies. We studied 304 twin individuals selected from the population-based FinnTwin16 study. Physical activity was assessed with the Baecke questionnaire, yielding three indexes: sport index, leisure-time index, and work index. In this study, we focused on sport index, which describes sports participation. Body composition was determined using dual-energy X-ray absorptiometry and cardiorespiratory fitness using a bicycle ergometer exercise test with gas exchange analysis. The Baecke sport index was associated with high maximal oxygen uptake adjusted for lean body mass (V̇o2max[adj]) (r = 0.40), with low body fat percentage (BF%) (r = −0.44) and low waist circumference (WC) (r = −0.29). Heritability estimates for the key traits were as follows: 56% for sport index, 71% for V̇o2max[adj], 77% for body mass index, 66% for WC, and 68% for BF%. The association between sport index and V̇o2max was mostly explained by genetic factors (70%), as were both the association between sport index and BF% (71%) and that between sport index and WC (59%). Our results suggest that genetic factors explain a considerable part of the associations between sports participation, cardiorespiratory fitness, and obesity.
- exercise participation
- maximal oxygen uptake
- genetic epidemiology
along with other health benefits, habitual physical activity prevents obesity (2, 11, 29) and increases cardiorespiratory fitness (4, 31). Previous studies have, however, shown that there are large individual differences in the response to physical exercise, which seem to be accounted for by genetic factors (4, 26).
Population-based twin studies have also shown that exercise behavior and obesity are strongly influenced by genetic factors. Heritability estimates of exercise participation ranged from 27 to 70% in a large pooled twin sample from seven countries (30), whereas the heritability of body mass index ranged from 45 to 85% in a study of twin cohorts in eight countries (27). For cardiorespiratory fitness, as measured by maximal oxygen uptake (V̇o2max), familial aggregation studies and twin studies show heritability estimates that vary between 50 and 67% (3, 8, 15).
Given the high heritability estimates, and the documented associations between physical activity, fitness, and obesity in observational studies (9, 10, 28), it can be hypothesized that obesity, exercise behavior, and cardiorespiratory fitness share some of their underlying genetic factors. However, the relationships between these traits are complex and presumably reciprocal, and few studies have tried to disentangle the genetic vs. environmental influences underlying these associations. Our laboratory's previous work (19) focused on these relationships for self-reported obesity measures and physical activity in the population-based FinnTwin16 cohort of young adult twins. We found that both additive genetic and unique environmental factors explained the inverse relationship between the obesity measures and physical activity. In contrast, the quantitative genetics of the relationships between objectively measured cardiorespiratory fitness and other traits have not, to the best of our knowledge, previously been studied. In the present study, we investigated the genetic and environmental relationships between objectively measured obesity, physical activity, and cardiorespiratory fitness in a subsample of Finnish young adult twins from the FinnTwin16 cohort.
The study participants were recruited from the FinnTwin16 cohort (13), a population-based, longitudinal study of five consecutive birth cohorts (1975–1979) of twins, their siblings, and their parents. We studied altogether 304 young adult twin individuals. They were selected by their responses to questions on weight and height at the age of 23–27 years to represent a wide range of intra-pair differences in body mass index (BMI). The sample was enriched with 20 monozygotic (MZ) and 53 dizygotic (DZ) pairs extremely discordant for BMI (intra-pair difference of >3 kg/m2) and 18 MZ and 13 DZ pairs concordant for BMI (intra-pair difference of <1 kg/m2) (EDAC: extremely concordant and discordant). Twenty-one MZ and 26 DZ pairs had intra-pair differences in BMI ranging from 1 to 3 kg/m2. Altogether, 59 MZ and 92 same-sex DZ pairs were studied. For two male subjects, data were available for only one member of a twin pair. Mean age of study subjects at the time of the clinical assessment was 27.4 years (standard deviation 2.0, range 23.3–31.5 years).
Body composition, including fat mass, lean body mass (LBM), and percent whole body fat (BF%) was measured by dual-energy X-ray absorptiometry (DEXA) (Lunar Prodigy, Madison, software version 2.15) (17).
BMI, defined as body weight in kilograms divided with the square of height in meters, was calculated from measured height and weight. Waist circumference (WC) was measured midway between the spina iliaca superior and the lower rib margin (16).
Cardiorespiratory fitness was measured by a work-conducted maximal exercise test with gas exchange analysis (spiroergometry) using an electrically braked bicycle ergometer, as described in an earlier report (18). The work load was initially 40 W for women and 50 W for men. It was increased by 40 W (women) or 50 W (men) every 3 min until the maximal exercise capacity was reached. Gas exchange including oxygen uptake (V̇o2) was measured breath-by-breath with a Vmax spiroergometer (Sensormedics, Yourba Linda, CA). The gas analyzers and the mass flow sensor of the device were calibrated before every exercise test. The electrochemical oxygen analyzer was calibrated with a gas mixture that had an oxygen concentration of 26% and the infrared CO2 analyzer with a gas mixture that had a CO2 concentration of 5% (Linde Gas, AGA, Espoo, Finland). The exercise was continued until exhaustion. The criteria for maximality of the exercise were rate of perceived exertion (RPE) 19–20/20 on the Borg scale or the gas exchange ratio V̇co2/V̇o2 of over 1.10. V̇o2max was defined as the mean V̇o2 measured during the last 30 s at the maximal work load. Maximal working capacity (Wmax) was defined as the mean workload during the last 3 min of exercise. In the analyses, we used V̇o2max adjusted by LBM (referred to as V̇o2max[adj]), since this seems the most unbiased way of comparing cardiorespiratory fitness of individuals of different sizes (6, 32). The same was done for Wmax (referred to as Wmax[adj]).
Physical activity was assessed using the Baecke questionnaire (1). It consists of 16 items, from which three indexes were calculated: work index refers to physical activity at work, sport index to sports participation during leisure time, and leisure-time index to physical activity during leisure time excluding sport activities. All responses were given on a five-point scale with the exception of questions on main occupation and the types of two main sports. In the original publication, the test-retest reliability of the work index, sport index, and leisure-time index were 0.88, 0.81, and 0.74, respectively (1).
Means and standard deviations were calculated for all variables. Before genetic modeling, Pearson correlations were calculated to investigate the relationships between the traits under study separately for men and women, whereas associations within twin pairs were studied with intraclass correlations for each trait, adjusted for sex. These analyses were performed using Stata (version 11) for Windows software (StataCorp, College Station, TX) and the Mx statistical package (20). The data were then analyzed using quantitative genetic modeling of twin and family data (21). Whereas MZ twins are genetically identical, DZ twins and full-siblings share, on average, 50% of their segregating genes identical-by-descent. In addition to additive genetic variation, which is the sum of the effects of all alleles affecting the phenotype, part of the genetic variation may be due to interaction between alleles in the same locus (dominance). Additive and dominance genetic effects are fully correlated within MZ pairs and have expected correlations of 0.5 and 0.25 within DZ pairs, respectively. Epistatic effects, i.e., interaction effects between alleles in different loci, are assumed to be absent. MZ and DZ pairs are assumed to share the same amount of environmental variation, which is partly shared by a twin pair (common environment) and partly unique to each twin individual (unique environment), including any random measurement error. Based on the above assumptions, four sources of variation interpreted as latent and standardized variance components in the structural equation model can be identified: additive genetic (A), genetic dominance (D), common environment (C), and unique environment (E).
Our data include only twins reared together and do therefore not allow modeling of genetic dominance and common environmental effects simultaneously. In the classical twin design, information on the variance components comes from three observed statistics: the phenotypic trait variance, the covariance between MZ twins, and the covariance between DZ twins. With three observed statistics, it is impossible to estimate four variance components (7). Because models that contain dominant genetic influences in the absence of additive genetic influences are biologically implausible (25), and because E effects include measurement error and as such have to be included in the model, a choice between ACE and ADE models has to be made. The genetic models were carried out using the Mx statistical package (20) using full information maximum likelihood. This implies that twin pairs with data available for only one twin are included in the model, since, although not informative of the covariances, they, however, provide information on the means and variances of the variables under study.
We started the genetic modeling by carrying out univariate models to calculate heritability estimates for all variables and to find the best models to be used for the variables selected for further modeling. These were chosen to best represent the traits of interest. In addition, the variables had to correlate to be included in the bivariate models.
The assumptions of twin modeling, i.e., equal means and variances for MZ and DZ twins as well as for both co-twins, were tested by comparing twin models to saturated models, which do not make these assumptions. Also, sex effects were tested in saturated models, and because significant mean differences between males and females were found in many variables, sex was included as a covariate in all models.
We then estimated pairwise genetic and environmental correlations between sports participation (represented by sport index), cardiorespiratory fitness (represented by V̇o2max[adj]), and obesity (represented by BF%, WC, and BMI) using bivariate Cholesky decomposition models. In these models, the relationship between two variables is modeled by decomposing the phenotypic covariance of the variables into proportions accounted for by A, C (or D), and E components. The degree of association of the additive genetic factors influencing the two variables can be estimated as the genetic correlation between the latent A factors for the two variables. Common and unique environmental correlations are estimated similarly. Sex was used as a covariate in all models because the sample size did not allow separate models for males and females.
In both univariate and bivariate modeling, the significance of each parameter in the model is tested by dropping the parameter and evaluating the change in −2 log likelihood between the initial model and the nested submodel. Model comparisons are made with likelihood ratio χ2 tests, where a significant change in χ2 indicates that dropping the parameter significantly decreases model fit, suggesting that the parameter should be retained in the model (21).
To study whether the selection procedures had any effects on the twin model estimates (5), we fitted univariate models for self-reported BMI using data from the full FinnTwin16 sample (1,532 MZ, 3,247 DZ twin individuals) and compared the estimates with measured BMI from the selected subsample. The heritability estimate from the best-fitting AE model of self-reported BMI in the full sample [A = 0.77; 95% confidence interval (CI): 0.74–0.80] was very similar to that for measured BMI in the selected sample (A = 0.72; 95% CI: 0.58–0.81), with clearly overlapping 95% confidence intervals.
The study subjects provided written, informed consent. The protocol was designed and performed according to the principles of the Helsinki Declaration and was approved by the Ethics Committees of the Helsinki University Department of Public Health and of the Helsinki and Uusimaa Hospital District.
MZ and DZ twins did not differ from each other on average for any of the body composition or physical activity measures. Of the cardiorespiratory fitness measures, only mean V̇o2max[adj] was slightly higher in MZ compared with DZ twins (Table 1). Males had significantly higher LBM, WC, and cardiorespiratory fitness than females but lower whole body adiposity and leisure-time activity index. BMI, sport index, and work index means did not differ between sexes (Table 1).
Phenotypic correlations for the main traits are shown in Table 2. Sport index was positively associated with V̇o2max[adj] and Wmax[adj] and inversely associated with the obesity measures. Wmax[adj] was inversely associated with the obesity measures, whereas correlations between V̇o2max[adj] and obesity measures were very weak or absent. In addition, there was a positive correlation between V̇o2max[adj] and leisure-time activity index in females but not in males, whereas a negative correlation between V̇o2max[adj] and work index was present in males only.
The within-pair intra-class correlations for all traits were higher for MZ twins than for DZ twins, indicating the probable effect of genetic factors on the traits studied (Table 3).
We started genetic modeling by estimating the best model for each trait. The additive genetic/specific environment (AE) model or the ADE model (with effects due to dominance) offered the best fit for all traits. Significant dominance effects were found for BMI (D = 0.74, P = 0.036), WC (D = 0.70, P = 0.041), LBM (D = 0.82, P = 0.001), fat mass (D = 0.64, P = 0.030), and BF% (D = 0.71, P = 0.047). However, the statistical power to distinguish between additive and nonadditive genetic effects was clearly insufficient, since the A effects were estimated at zero in all of these ADE models. Dominance effects in the absence of additive genetic effects are biologically implausible (25, 33). Thus, in the subsequent modeling, we decided to use the AE model with sex as covariate, since sample size did not allow separate models for males and females.
Heritability estimates from univariate models (Table 3) ranged from moderate to high and were higher for body composition and cardiorespiratory fitness than for physical activity. Based on the univariate models, we used bivariate AE models to estimate the relative importance of genetic and environmental factors on the relationships between sports participation (sport index), cardiorespiratory fitness (V̇o2max[adj]), and obesity (BF%, WC, and BMI) (Table 4). Of the physical activity indexes, sport index was chosen, since it was the only one associated with both the cardiorespiratory fitness and the obesity measures in both sexes. V̇o2max is commonly used as a measure of cardiorespiratory fitness in the literature. Thus we chose to use it rather than the closely correlated but rarely used Wmax.
In the bivariate models, we focused on decomposing the phenotypic correlation between traits into genetic and environmental components. According to these models (Table 4) additive genetic factors accounted for a major part of the relationship between sport index and V̇o2max[adj] and of that between sport index and BF% (70 and 71%, respectively). Also, environmental correlations were significant in both cases and could not be dropped from the models [sport index and V̇o2max[adj]: Δχ2 = 5.34, degrees of freedom (df) = 1, P = 0.02; sport index and BF%: Δχ2 = 5.16, df = 1, P = 0.02]. Figure 1 offers a schematic presentation of the modeling results for sport index, V̇o2max[adj], and BF%. Likewise, the correlations between sport index and both WC and BMI were predominantly explained by genetic factors (Table 4). In the bivariate model for sport index and BMI, the specific environmental correlation (rE) was not statistically significant and could thus be dropped (Δχ2 = 2.50, df = 1, P = 0.11), whereas, in the model for sport index and WC, the rE was significant and was retained in the model (Δχ2 = 5.38, df = 1, P = 0.02). In Table 4, we present the results of the full models as well as the model for sport index and BMI from which rE has been dropped. The relationship between V̇o2max and obesity was not modeled, because their phenotypic correlation was small and not statistically significant.
In the present study, we studied the relationships between body composition, sports participation, and cardiorespiratory fitness in young adult twins from a Finnish twin cohort. We found a strong negative correlation between obesity, as measured by body fat percentage and waist circumference, and sports participation, as measured by the Baecke sport index. Sport index was, in addition, positively associated with cardiorespiratory fitness, as measured by V̇o2max adjusted for LBM. Obesity and cardiorespiratory fitness were not, however, associated.
In bivariate genetic analyses, we found that genetic effects accounted for ∼70% of the covariance both between sports participation and obesity (as measured by BF%) and between sports participation and cardiorespiratory fitness (as measured by V̇o2max[adj]). This indicates a major role of genetic factors in the relationships between vigorous physical activity and traits associated with it. Exercise participation itself is a heritable trait (30), as are obesity (27) and cardiorespiratory fitness (3, 8, 15). This study shows that there are shared genetic factors behind these traits and that the relationships between vigorous physical activity and both V̇o2max[adj] and BF% are in large part genetically mediated.
Weaker correlations were seen between sport index and BMI, and sport index and WC, compared with that between sport index and BF%. This may indicate that BF% is a better measure than BMI and WC of overall obesity in young adults. The aforementioned obesity measures reflect different aspects of body composition, which is likely to cause the differences seen in the genetic contribution on the relationships between sport participation and these measures.
Our result emphasizes the role of genetic factors in the relationship between exercise and cardiorespiratory fitness, which is in line with previous studies showing large individual differences in physiological responses to physical activity. In the HERITAGE family study, Bouchard et al. investigated the heritability of exercise-training-induced changes in several phenotypes, such as V̇o2max, submaximal aerobic performance, hemodynamic phenotypes, body composition, and body fat distribution. Heritabilities for these traits ranged from 25 to 55%, indicating a major role of genetic factors in the responsiveness to physical activity (26). The specific genes involved in the interplay between physical activity and the exercise-related phenotypes are still poorly known, as are the mechanisms by which they influence response to training. A number of genes have, in single studies, been found to be associated with change in body composition and V̇o2max in response to exercise training, but none of these results have been replicated (26). In genetic studies seeking to identify specific genes, the present study may provide information about combinations of variables or bivariate models useful in identifying meaningful phenotypes for analysis.
We also found, somewhat surprisingly, that although sports participation was associated with both V̇o2max[adj] and BF%, these traits were not associated with each other. There are several possible explanations to this. One is that since the metabolic pathways by which physical exercise enhances fitness and reduces body fat are different, these may be influenced independently of each other, and the responses to physical activity may be different in individuals with different genotypes. In other words, the individuals who respond to physical activity by increasing their V̇o2 capacity are not necessarily the same who reduce their amount of body fat. Another explanation is that individuals who seem to be on the same exercise level by the Baecke sport index might in fact exercise quite differently when it comes to frequency and intensity of the training. Rather short and nonregular but intensive exercise sessions may, in fact, improve cardiorespiratory fitness without reducing obesity (22).
The sample used in the present study was partly selected on the basis of pairwise discordance and concordance for obesity (known as EDAC selection), as assessed by BMI calculated from self-reported weight and height. This selection procedure yielded a highly informative sample for studying the relationships between cardiorespiratory fitness, exercise behavior, and adiposity. Concerning the representativeness of the twin model estimates from this selected sample, it has been shown with simulated twin data that the bias resulting from the EDAC selection is minimal (5). Because data selected with the EDAC procedure are technically “missing at random,” unbiased model estimates are in fact expected on the basis of missing data theory (14). In the present study, the heritability of BMI in the subsample was very close to that of the full sample and in agreement with heritability estimates derived from earlier twin studies on young adults (27).
This study has several strengths. Studying MZ and DZ twins makes it possible to examine both genetic and environmental influences on certain traits as well as their interactions. Furthermore, we have used state-of-the-art methods to measure body composition and cardiorespiratory fitness in the study subjects. Physical activity was assessed with the well validated Baecke questionnaire (12, 23, 24), which, in a study investigating the validity of three commonly used physical activity questionnaires, showed the highest correlation with physical activity level, as measured with the doubly labeled water method (24). The main limitation of the study was the relatively small sample size, which reduced the power of twin modeling to distinguish between genetic and environmental relationships between the variables, especially when phenotypic correlations were weak. In the absence of opposite-sex pairs, sex-specific genetic effects could not be investigated, and the sample size did not allow modeling sexes separately. To compensate for this, we included sex as covariate in all models.
The present results show that higher sports participation is associated with higher cardiorespiratory fitness and less overall and abdominal obesity and that these relationships are to a great extent explained by genetic factors.
This study was funded by the Academy of Finland Center of Excellence in Complex Disease Genetics, Academy of Finland (grant nos. 44069, 100499, 118555, and 108297) and DIOGENES (“Diet, Obesity, and Genes”) project supported by the European Union (contract no. FP6–513946), the European Community's Seventh Framework Program (FP7/2007-2013), ENGAGE Consortium, grant agreement HEALTH-F4-2007-201413, and Helsinki University Central Hospital grants. L. Mustelin was supported by the Research Foundation of the University of Helsinki.
No conflicts of interest, financial or otherwise, are declared by the author(s).
- Copyright © 2011 the American Physiological Society