The heritability of physical fitness characteristics has been widely investigated. Most early genetic studies of physical fitness concentrated on performance measures such as static strength, explosive strength, and running speed, and some included tests for speed of limb movement and balance (28). Recently the relationship of fitness and health has gained attention (1) and has led to the study of other, health-related fitness components, especially cardiorespiratory fitness and adiposity. Few studies report on the genetic determination of flexibility and functional or trunk strength. The most striking conclusion of a comparison of the studies is the large variability in heritability estimates with age, gender, population, and methods used to estimate them (5,28). This variability in the reported values of genetic predisposition is partly owing to deficiencies in the design and analysis of the studies. Sample sizes are often very small, untenable assumptions are made, zygosity is often poorly determined in twin studies, and proper adjustments for age, gender, and effects of other concomitant variables, especially in familial studies, are not made(5). The estimates of genetic determinants for fitness components are almost always higher when derived from twin studies than those derived from family studies in which parent-child and/or sibling correlations are calculated (6,20).
The heritability estimates for static strength obtained from twin studies vary between 0.0 and 0.89(13,18,24,25,38). Coefficients derived from family studies vary between 0.09 and 0.61(30,35-37,46,52). Most of these studies report moderate genetic variance for static strength. Explosive strength as measured by vertical jump is more strongly genetically determined. Coefficients vary between 0.82 and 0.93 for twin studies and between 0.22 and 0.68 for family studies(13,25,29,35,38,46,51,52). Data for other performance-related fitness components such as running speed, speed of limb movement, and balance are limited and also show large variability (28). The genetic predisposition of cardiorespiratory fitness performance is particularly equivocal. Some(13,23,24,38,41,51) conclude that maximal oxygen uptake (˙VO2peak) is almost entirely determined by heredity, while others(7-9,26,27,37,40) state that less than half of the variation in ˙VO2peak is genetically determined. While the older twin studies report very high heritability estimates (0.80-0.95), recent twin studies present much lower ones(0.40-0.70). In agreement with these latter studies, family studies suggest a genetic effect of between 0.30 and 0.50 for cardiorespiratory fitness based on familial correlations. Generally, parent-child correlations (0.06-0.34) are lower than sibling correlations (0.14-0.43). Few studies report on the heritability of functional strength and trunk strength. Twin data suggest that the two health-related strength factors are less influenced by genetic factors(0.20-0.70) than the more performance-related strength factors (static and explosive strength) (0.60-0.90) (25). For the estimates based on family data, the trend is less obvious(35,36). Flexibility was rarely investigated from a genetic perspective (15,36) and was found to be moderately to highly heritable (0.45-0.90). Adiposity, however, has been the object of many quantitative genetic studies, including four studies that used path analysis and structural equation modeling(3,4,11,12). Most studies that evaluate adiposity by measuring skinfold thickness report a significant genetic component, ranging mostly from 40-70%. Compared with other anthropometric measures often included in these studies, skinfolds are among the least heritable. Environmental effects are suggested as an explanation for the lower parent-offspring and sibling correlations for subcutaneous fat(21). Only recently has structural equation modeling been used to quantify genetic and environmental sources of variation in fitness characteristics. Pérusse et al. (36) reported estimates of transmissibility between 0.28 for physical working capacity (PWC) relative to body mass and 0.48 for flexibility. The transmisibility estimates reported by Devor and Crawford (15) ranged between 0.0 for hand strength and 0.66 for trunk flexibility. Based on the results of these studies, it was hypothesized that performance-related fitness characteristics are more under genetic control than are healthrelated fitness items.
The present study focuses on the quantification of genetic and environmental sources of variation in physical fitness components observed in 105 twin pairs and their parents (97 mothers and 84 fathers) using structural equation modeling. The advantages of this study are: 1) a large population sample of both male and female twins together with their parents, 2) fixed twins' age (10 yr), 3) good measures of performance-related and health-related fitness, 4) good statistical methods using all families even those with missing data, giving maximum likelihood estimates of the contribution of genes and environment to fitness(17,22,32,33). More specifically, the following hypotheses are tested: 1) broad heritability is higher for performance-related fitness than for health-related fitness; 2) common environmental factors play a significant role in health-related fitness; 3) dominance does not contribute significantly to variation in physical fitness; 4) genetic variance is higher for girls than boys for some traits with early peak velocity; 5) different genes influence fitness in parents and offspring for the same traits; 6) the contribution of cultural transmission is moderate at this age; and 7) assortative mating increases the genetic variance for some traits.
Subjects. In a longitudinal project (Leuven Longitudinal Twin Study) a variety of somatic and physical fitness data were collected from 105 twin pairs and their parents. This sample represents an approximately 50% response rate to a recruitment procedure by telephone and/or home visit. The twins first participated in the fitness testing at a prepubertal stage (at age 10) and will be followed up annually until after puberty, with semiannual observations for the anthropometric characteristics. The data reported here concern the first visits of 105 twin pairs and their parents. These 10-yr-old twin pairs ([horizontal bar over]X = 10.3, SD = 0.3) were approximately equally divided into the five twin groups: (21 male monozygotic pairs (MZMM), 22 female monozygotic pairs (MZFF), 21 male dizygotic pairs (DZMM), 19 female dizygotic pairs (DZFF), and 21 unlike-sexed dizygotic pairs (DZMF)). Data were available for 95% of the mothers (N = 97) and 90% of the fathers(N = 84). All subjects were informed about the study, its longitudinal character, and the tests and measurements done. The project was approved by the Medical Ethics Committee of the Fund for Medical Research and of the Institute of Physical Education, K. U. Leuven.
Determination of zygosity. All twins were selected from the East Flanders Prospective Twin Study (14). In each twin gestation the fetal membranes were examined and placental morphometry was performed. Placental alkaline phosphatases were assayed by electrophoresis, and umbilical cord blood was used to determine the ABO, Rh, MNSs, Duffy, and Kell blood groups by routine methods. Also DNA restriction fragment length polymorphisms were studied. Zygosity was determined through sequential analysis (50). Unlike-sexed twins and same-sexed dichorionic twins with at least one different genetic marker were classified as dizygotic. Monochorionic twins were classified as monozygotic. The probability of monozygosity based on the genetic markers was calculated for all same-sexed dichorionic twins with identical genetic markers. All dichorionic twins of the same sex and same markers, reaching a probability of monozygosity of 0.90 or more, were considered monozygotic.
Variables. Selection of the performance-related fitness items was based on the factor analytic studies of Simons et al.(43,44) that resulted in the construction of the Eurofit test battery. Five performance-related fitness characteristics were examined: static strength (arm pull), explosive strength (vertical jump), running speed (shuttle run), speed of limb movement (plate tapping), and balance (flamingo balance). The four health-related fitness characteristics were flexibility (sit and reach), trunk strength (leg lifts), functional strength (bent arm hang) (both measures of local muscle endurance), and maximal oxygen uptake ((˙VO2peak) as an indicator of cardiorespiratory fitness) measured during a maximal exercise test on a treadmill. A description of the tests is given in Simons(45) and by Bruce (10) for the treadmill run. The test-retest reliability varied between 0.65 and 0.96(34,45). A fifth fitness component, adiposity, was evaluated by the sum of six skinfold measurements taken at representative sites of the body (biceps, triceps, suprailiacal, subscapular, medial, and lateral calf) according to the procedures described in Simons(45).
Statistical analysis. To determine the relative contribution of genetic and environmental factors on the variation of fitness characteristics, path models were fitted to the data(17,22,32,33). First, some assumptions of these models were tested, including a test for the normal distribution(Shapiro-Wilk test) and significance tests for differences in birth order, zygosity, and gender (by fitting models to the means and covariances of the data by maximum likelihood). Second, maximum likelihood estimates of the correlations between the family members were computed based on the raw data using Mx (31) to apply equality constraints on certain correlations. For example, the correlation between father and first-born twin is expected to be equal to that between father and second-born twin. The ratio of the MZ and DZ correlations for males and females is informative for the parameters to be included in the genetic models.
Variance-covariance matrices were used as input statistics for the twin models. The twin-parent design was fitted to the raw data. Mx, a structural modeling equation package (31), was used to compute the goodness of fit of the models and maximum likelihood estimates of the parameters.
In a first analysis only the twin data were used, whereas in the second analysis both the data of twins and their parents were simultaneously entered in the models. We present the twin analyses because 1) the twin models are simple and basic, 2) they can be compared with other twin studies, and 3) they can be compared with the twin-parent findings here. We then present the twin-parent analyses because 1) they give better estimates taking into account assortment and cultural transmission, 2) they offer a test of the twin study assumptions, and 3) they can be compared with other parent-offspring studies. In these linear models the causal paths (between the latent and observed variables) and the correlational paths (between the latent variables) are specified. Possible causes of variation are additive genes (A), dominance (D), common environment (C), specific environment (E), and phenotypic interaction(P).
With data on MZ and DZ twins reared together, the significance of additive genetic variance, specific environmental factors, and either a common environmental or a dominance parameter can be tested. For the twin data alternative models were tested for all fitness items (Fig. 1). Starting from the ACE-model (33, p. 98) and assuming additive genetic (A), common environmental (C), and specific environmental (E) factors, three submodels (AE-model, CE-model, E-model) were fitted to the data. The AE model tested the significance of the shared or common environmental factors (c2). The CE model tested whether additive genetic factors (a2) could be dropped from the full model. Whether the observed data could be explained purely by random environment (e2) without genetic and familial resemblance was tested in the E-model. The common environmental factor was replaced by one accounting for variance owing to dominance (d2), the ADE-model (33, p. 153). Since the effects of shared environmental factors and genetic dominance are confounded in studies of twins reared together, a full model including additive and dominance genetic factors and specific and shared environmental factors (ACED) cannot be tested. Finally, the presence of phenotypic interaction (i2) was verified starting from the ACE-model (PACE-model) or from the AE-model(APE-model) (33, p. 201). These models included a parameter accounting for the direct effect of one twin's phenotype on his cotwin's phenotype and vice versa. These alternative models were compared by the chi-squared goodness-of-fit statistics. Akaike's Information Criterion (AIC) and likelihood ratio tests were used to indicate the most parsimonious and best fitting model(22,32,33). Although, while using AIC, a particular model may be preferred, this model may not have a significantly lower chi-squared than the simpler model.
A heterogeneity analysis was performed to test whether the genetic architecture of fitness characteristics was different in males and females. Models including additive genetic effects and common and specific environmental factors were fitted to the data with parameters alternatively set equal or allowed to differ for both genders (33, p. 212). The full scalar sex limitation model tests whether the magnitude of the genetic and environmental factors is different for males and females (S-AE). Models in which some parameters are different between sexes while others are equal (e.g., SE-AE: different specific environmental parameters for males(em) and females (ef) but equal genetic parameters for both genders) or models with a different set of parameters for males and females (e.g., an AE-model for males and a CE-model for females(AE/CE)) are submodels of the full scalar sex limitation model. A difference in total variance in males and females may result in a general scalar difference of the genetic and environmental estimates, which was tested in the general scalar sex limitation model (GS-AE). Finally, a different set of genes or common environmental factors may act in males and females. These models are referred to as the nonscalar sex limitation model (NS-AE) or the general sex limitation model.
Including the fitness characteristics of the parents with the twin design not only increases the power of the design but also allows the testing of a variety of models and assumptions (17,20). In addition to the genetic and environmental parameters, the significance of new parameters was tested in different submodels. The shared environment parameter may consist of the impact of the parents' phenotypes (cultural transmission,z) and of shared environmental parameters independent of those transmitted through the parents (nonparental shared environment, b)(see Fig. 2). With data of twins and their parents, a model involving both cultural transmission and genetic transmission is identified. The result of the combined presence of genetic and cultural transmission in natural families is covariance between the genetic and shared environmental factors in the parents and the off-spring. It is assumed that this genotype-environment covariance (s) is at equilibrium over generations. Furthermore, the covariance between husband and wife(m) allows us to model assortative mating. To do this we assume that mate selection is based on the phenotype, which generates covariances between the genetic, shared, and specific environmental latent variables of the parents. This model has been used in many other applications(Fig. 2)(17,20,33). Covariance between spouses can also result from spousal interaction, but such effects cannot be separated from assortment with twin-parent data on one occasion of measurement. In the absence of cultural transmission, the only relevant covariance between the parents is then the covariance between their genetic factors. As the twins were all measured at age 10, but the parents as adults, we fitted additional models that did not constrain the genes operating in adults to be the same as those operating in 10-yr-old children. This reduced genetic transmission was obtained by freeing the genetic transmission parameter (g) to values between 0 and 0.5 (Fig. 3). This model has the undesirable assumption that the heritability is the same in both generations, that is, different genes play a role in children and adults, but the magnitude of their effect is the same. Data from adult twins are necessary to relax this assumption.
Testing for gender heterogeneity in the twin-parent model is not a straightforward extension as in the case for the twin model. A model allowing for sex differences involving more than one generation was conceived by Eaves et al. (16). For this study this model was implemented in Mx and adapted to fit to raw data of twins and their parents. A full description of this model is beyond the purpose of this paper, but the main characteristics are highlighted. The extension of the twin-parent model to test for scalar sex limitation is fairly simple and mainly involves estimating different parameters in males and females. To implement a nonscalar sex limitation model, however, involves estimating the correlation between additive genetic (or common environmental) latent factors in males and females. The expected correlation between the additive genetic factors of DZ twins is derived through genetic transmission from the parents. Each child inherits half his genes from each parent; therefore, the genetic transmission parameters are fixed to 0.5, regardless of the sex of the parent and the offspring. While males may express different genes than females for a specific trait, they still inherit half their genes from their mother. The genetic correlation between the sexes can be estimated in a Cholesky decomposition. Let us assume a set of genes common to both sexes (A) and a set of genes specific to one sex (B). (Estimating two sets of gender-specific genes is not identified; therefore, either a male-specific or a female-specific set is estimated). Both sets of genes are transmitted from parents to offspring. The gender-common set is expressed in both sexes and the male-specific set is only expressed in males (Fig. 4). A similar Cholesky decomposition is applied to common environmental effects (C and D) and dominance factors (K and L). Likewise, cultural transmission parameters are gender common (m,p) or male specific(n,o) and of maternal or paternal origin. Unique environmental factors are estimated separately for males (E) and females (F). The twin-parent design does not allow simultaneously testing of shared environmental effects and dominance. The model with full sex differences includes 30 parameters, including three parameters for the means, two variance parameters, and 12 constraints. The raw data are read in for each of the five zygosity/sex groups from variable length (VL) files (31, p. 14). Heterogeneity testing is done for either the cultural transmission or the dominance model. Depending on the best fitting model of this analysis, further parameters are dropped to test their significance. The most parsimonious model was selected by AIC, which takes into account both goodness-of-fit and degrees of freedom (48).
Twin data analysis. The representativeness of the twin sample was tested by comparing the data of the twins and parents, summarized by gender and zygosity in Table 1, to reference values of the Flemish population (2,45). The only significant differences were for balance and running speed in girls and for adiposity in both sexes. Both twin girls and twin boys have significantly less subcutaneous fat than singletons. Fathers of the twins scored significantly less on all the fitness tests, except for static strength, and showed more trunk adiposity than a 35-yr-old reference group. These differences can probably be accounted for by the age range (30-60 yr) of the fathers. It was therefore concluded that the twin sample was representative of the Flemish population for the fitness items considered.
The descriptive statistics indicate that most fitness items depart from the normal distribution. Only the extreme skewness for functional strength (bent arm hang) and adiposity (sum of skinfolds) was corrected by logarithmic transformation. The comparison between the means and variances of the twins grouped according to gender, zygosity, and birth order revealed gender differences in most fitness characteristics. Boys outperformed girls in arm pull, vertical jump, bent arm hang, and the treadmill run, whereas girls scored better for flamingo balance and flexibility. Girls had significantly more fat than boys, showing 10-30% larger skinfolds. The variance of the sum of skinfolds was also significantly larger in girls than boys. Comparing MZ and DZ twins within each gender revealed a greater static strength (arm pull) for MZ boys and a better flexibility (sit and reach) for MZ girls, whereas the variance of DZ boys was greater for speed of limb movement (plate tapping) and leg lifts (trunk strength). The assumption of equal means and variances by birth order and zygosity was, however, met for most variables. Given the large number of tests, some significant results would be expected to occur by chance.
Table 2 presents the combined MZ and DZ twin correlations and the correlations for male and female MZ and DZ twins and opposite-gender twin pairs separately. For most fitness tests the MZ correlations were twice as high as the DZ correlations, indicating a genetic effect. Balance and plate tapping showed a different pattern, with DZ correlations almost as high as the MZ correlations, suggesting that most of the variance is nongenetic in origin. The DZ correlations for running speed, flexibility, and adiposity were greater than half the MZ correlations, indicating the possibility of shared environmental factors contributing to the variance, besides additive genetic factors. The twin correlations split up by gender give an indication of the heterogeneity of the genetic determination. It should be noted that the sample sizes were small and that consequently the standard errors of these correlations were high. The gender specific MZ correlations exceeded DZ correlations except for cardiorespiratory fitness in males, which suggests genetic influences. The male DZ correlation was markedly less than half the MZ correlation for arm pull, possibly indicating the existence of genetic dominance. On the other hand, the DZ correlations for speed of limb movement and flexibility in both genders, for static strength, running speed, and trunk strength in females, and for balance and cardiorespiratory fitness in males were markedly greater than half of their respective MZ correlations. This may be a result of common environmental factors or the genetic effects of assortative mating. The opposite-sex correlations were lower than the same-sex correlations for explosive strength, flexibility, and maximum oxygen uptake, suggesting that different sets of genes have an effect in the two genders.
Heterogeneity models were fitted to the twin data in five gender- and zygosity-specific groups. In Table 3 the most parsimonious model and the corresponding chisquared, degrees of freedom, probability, and AIC were reported together with the percentages of explained variance by each of the parameters included in the model. Standardized parameter estimates are given by sex in case of evidence of gender heterogeneity; otherwise they are only listed in the “male” column. The best fitting model for most fitness items was the AE model. Dropping the genetic factors resulted in a marked decrease of fit. On the other hand, adding parameters for genetic dominance or phenotypic interaction did not significantly improve the fit of the models. Including common environment (ACE model), however, resulted in a significantly better fit for running speed and flexibility. Genetic factors accounted for 23% (running speed) to 72% (static strength) of the phenotypic variance in the performance-related fitness scores and for 38% (flexibility) to 86% (adiposity) of the phenotypic variance in the health-related fitness scores. Between 32% and 46% of the variance in flexibility and running speed was a result of shared environmental factors. The remaining variance was explained by specific environmental factors. Although the fit of the AE model was good for most items, the significance of the chi-squared was borderline for trunk strength. This poorer fit could be due to the difference in the total variance of MZ and DZ twin pairs.
No sex differences in genetic architecture were found for most strength factors (static, functional, and trunk strength), speed of limb movement, and balance. The AE-model with equal parameters for males and females remained the best fitting model for these characteristics. Genes account for about 70% of the strength factors and 50% of speed of limb movement and balance. For explosive strength a model with additive genetic and specific environmental parameters also gave the best fit, but gender heterogeneity was found for both genetic and specific environmental components. In females the genetic factors explained 78% of the variance, whereas in males genetic factors explained only 47% of the variance. For running speed and flexibility an ACE-model resulted in the best fit with gender heterogeneity in the environmental component. The genetic component accounted only for 23% of the total variance in males and 33% in females for running speed. Common environmental factors explained 33% in males and 46% in females, while specific environmental factors accounted for 44% in males and 21% in females. For flexibility the genetic factors explained 50% of the variance in females and 38% in males. The remaining environmental variance was mostly shared in female twins and equally divided in a shared and nonshared environment for males. For the sum of skinfolds, on the other hand, the parameter for the genetic factors differed in males and females. The genetic determination of skinfolds was greater in 10-yr-old females (86%) than in males (80%). For cardiorespiratory fitness, no evidence was found for genetic factors in males, while 85% of the variance was explained by genetic factors in females. The reverse was found for common environment, which accounted for 66% of the variance in males and 0% in females.
Twin-parent data analysis.Table 4 includes both father-child (FC) and mother-child (MC) correlations, the parental or marital correlation (FM), and parent-offspring correlations specific to the gender of the child such as the father-son (FS), mother-son (MS), father-daughter (FD), and mother-daughter (MD) correlations. No significant differences were found between father-child and mother-child coefficients. The parent-child correlations by gender of the child showed no clear trends, but they were generally lower than the DZ twin correlations. It is thus not likely that positive cultural transmission will be significant. This pattern is consistent, however, with dominance or genetic heterogeneity across generations. Highly significant husband-wife correlations (0.26-0.48) were found for running speed, speed of limb movement, balance, trunk strength, and maximum oxygen uptake, suggesting assortative mating for these characteristics. The marital correlation for adipositiy was also significant but negative (-0.19).
The model fitting results obtained in the twin analyses were corroborated by the twin-parent analyses (Table 5). Additive genetic factors and specific environmental factors accounted for most of the variance in fitness characteristics. No evidence was found for positive cultural transmission. Common environmental effects of nonparental origin were only significant for the speed measures (running speed and speed of limb movement) and flexibility in females. Since the DZ correlation was greater than half the MZ correlation and exceeded the parent-child correlations, nonparental shared environmental effects rather than cultural transmission for these measures were significant. Twins seem to share environmental factors with their cotwin but not with their parents. Other mechanisms may be responsible for lower parent-child versus DZ correlations. First, genetic dominance effects correlate 0.25 in DZ twins, but 0.0 in parent-offspring pairs. Second, parents may negatively influence their children (“negative cultural transmission”). Third, different genes may influence characteristics in children and in adults. Power to discriminate between these three models is expected to be very low. This is illustrated by the close fit of these models for static strength, explosive strength, functional strength, maximum oxygen uptake, and adiposity. The three models estimate the specific environmental variance almost identically. The remainder of the variance is either split by additive and dominance variance, large negative genotype-environment covariance, or additive genetic variance with a low genetic correlation between generations. There is evidence of assortative mating for the speed measures, balance, trunk strength, and maximal oxygen uptake. The most parsimonious models are thus the ACE-model for flexibility, the ACEM-model for running speed and speed of limb movement, the AEM-model for balance and trunk strength, the ADE/AZE/AEG-model for static, explosive, and functional strength, the ADEM/AZEM/AEMG-model for maximum oxygen uptake, and the AEM/AZEM for adiposity. Since negative cultural transmission does not seem plausible for the traits under study, estimates from either the dominance or reduced genetic transmission model will be presented.
For the performance-related fitness scores, the estimated explained variance in the twin and parent data (Fig. 5) owing to the genetic component was highest for static strength (74%). This percentage of explained variance was the same as in the twin data analysis(Table 3). While the heritability for explosive strength was estimated differently for males and females in the twin analysis, no evidence for sex differences in broad heritability (65%) was found in the twin-parent model. A dominance/reduced genetic transmission model fitted these two variables best. The observed twin and parent-offspring correlations can be explained either by 34-40% dominance variance or by a genetic correlation between generations of 0.56 for static strength or 0.38 for explosive strength. For the speed of limb movement, the heritability estimates were lower in the twin-parent analysis than in the twin data analysis owing to the significance of nonparental shared environmental effects shown only in the twin-parent analysis. This finding illustrates that the twin design is not very powerful in detecting common environmental factors. However, as indicated by running speed, common environmental variance of about 30% is detectable with a sample size of 105 twins. According to the twin-parent data, the genetic component explained 23-41%, the common environment explained 23-37%, and the specific environment 22-54% of the total variance for both speed measures. Sex differences remain significant for running speed. For balance the specific environmental component explained the majority of the variance(59%). The remainder of the variance is additive genetic. The genetic variance for the speed measures and balance is increased by positive assortative mating by 2-7%.
Twin-parent model fitting results are fairly similar to the analyses on twin data alone for two of the health-related fitness items: flexibility and trunk strength. For flexibility, common environmental variance was corroborated in girls but not in boys. Additive genes explained 72% of the variance for males and 51% for females. The heritability for trunk strength was estimated to be slightly lower than in the twin analysis. The genetic variance was 63% of which 10% resulted from positive assortative mating. The decision for the best fitting model was not straightforward for functional strength and cardiorespiratory fitness. The goodness-of-fit was similar for a dominance model, a cultural transmission model with negative genotype-environment covariance, and a reduced genetic transmission model. Both the dominance and the reduced genetic transmission model estimated the broad heritability at 77% for functional strength and at 69% and 87% for males and females, respectively, for cardiorespiratory fitness. Allowing for different genes to be operating in children and adults (rg = 0.31), the percentage of explained additive genetic variance remained the same as the twin analyses for functional strength. The same holds true for cardiorespiratory fitness in girls but not for boys. A common environmental explanation was preferred for boys in the twin analysis. The twin-parent analysis, however, favored a genetic explanation with either dominance or reduced genetic transmission. The genetic correlation between generations was estimated at 0.74, and a different set of genes explained the variance in males and females. A nonscalar sex limitation model also fitted slightly better than a cultural transmission model for adiposity. According to this model 79% of the variance in males and 90% of the variance in females is explained by additive genetic factors, a result that is very similar to the twin analysis. Taking into account assortative mating further affected the genetic variance for both cardiorespiratory fitness and adiposity between 2% and 6%.
To quantify the contribution of genetic and environmental factors to the variation in fitness, genetic models were fitted to data from 10-yr-old twins and their parents. In contrast to previous analyses, the answer is not always straightforward. The primary hypothesis that performance-related fitness is more heritable than health-related fitness was not answered in a clear-cut way. Overall, more evidence was found for the reverse; namely, the range of heritability estimates for health-related fitness (51-90%) is higher than for performance-related fitness (23-74%). Our analyses confirmed previous studies showing that most fitness characteristics have high heritabilities and that a simple additive genetic and specific environmental model (AE) explained most of the variance. Few of the previous analyses go further than estimating the broad heritability. A better understanding of the fitness phenotypes may be obtained by testing for the significance of shared environment and for dominance and assortative mating. The components of variance may also differ according to sex and age. The classic twin design allows testing for some of these factors, including dominance and common environment, as well as phenotypic interaction by fitting alternative structural equation models. Several sex limitation models can account for gender heterogeneity. If the twin design is augmented with data from their parents, a more complete picture may be drawn. Besides cultural transmission, genotype-environment interaction, and assortative mating, a simplified model for different gene action in consecutive generations may be tested. More complex patterns of sex differences in the parent-offspring design may be evaluated.
Specific hypotheses were formulated to test for a variety of possible mechanisms. The first hypothesis that common environmental factors play a more significant role in health-related fitness was not supported at all. In fact, the common environment was only significant for flexibility in girls and for speed measures. Although shared environmental factors seemed important for cardiorespiratory fitness in boys in the twin analysis, it was not confirmed in the twin-parent analysis. Secondly, no evidence was found for phenotypic interaction for any of the fitness scores. Dominance appeared to be one possible explanation for the pattern of covariances of parents and offspring for static, explosive, and functional strength, and for cardiorespiratory fitness. However reduced, genetic transmission was almost as likely. If the dominance model is selected, dominance variance explained between 34% and 52% of the variance.
Sex differences in heritability were observed in both analyses for running speed, flexibility, cardiorespiratory fitness, and adiposity. These fitness characteristics show an earlier peak growth. Assuming that maturity is highly genetic, the gender difference may reflect the advanced maturity of girls over boys at this age. The hypothesis that heritability would be greater in girls than in boys was rejected only for flexibility. The difference in total variance for flexibility between genders was best explained by a significant common environmental effect in girls only. Nonscalar sex limitation seems a more likely explanation for the variance in adiposity than negative genotype-environment covariance and cultural transmission in girls.
The hypotheses of developmental change, cultural transmission, and assortative mating could only be tested by including the parents in the design. First, heterogeneity between generations for gene action was as good an explanation as dominance or negative genotype-environment covariance for the low parent-offspring correlations for three strength factors and cardiorespiratory fitness. These results, however, need to be replicated by including data on adult twins, which will allow relaxing the constraint of equal heritability in both generations. Second, as expected, no evidence was found for cultural transmission except as an unlikely explanation for variance of strength and cardiorespiratory fitness. For the traits with evidence for common environmental variance, it is expected to be of nonparental origin. Third, assortative mating, although modest for most variables, increased the genetic variance by 2% to 7% for speed measures, balance, trunk strength, and cardiorespiratory fitness. However, a negative husband-wife correlation was observed for adiposity.
To our knowledge, only three studies(15,19,36) have used comparable statistical methods to assess the degree of familial transmission in physical fitness, with the exception of adiposity. Family studies by Pérusse et al.(36,37) focus on the intergenerational transmission of several physical fitness characteristics, including grip strength, push-ups, sit-ups, trunk flexion, and physical working capacity(PWC150). Grip strength corrected for body weight and trunk flexibility were also included by Devor and Crawford (15). Both push-ups and sit-ups are assumed to measure trunk strength; trunk flexion is a flexibility test and PWC150 is an indicator for cardiorespiratory fitness. Aerobic power (˙VO2peak), among other cardiac parameters, was investigated by Fagard et al. (19).
Pérusse et al. (36) found a total transmissibility (both genetic and cultural) of 0.37 whereas Devor and Crawford (15) found no transmissibility for handgrip strength. Pérusse et al. (36) also found significant positive assortative mating for static strength. These results do not correspond well with our findings for the arm pull test. However, in the present study, the absolute value of arm pull strength was considered, not a relative score expressed as per kilogram body weight. In recent studies(36,42) of grip strength, maximal aerobic power, anaerobic capacity, muscle fiber type composition, and enzyme activities, the performance scores are often corrected for body size. For example, common practice is to divide ˙VO2peak by body weight or, when available, by fat free body weight. In our analyses we used the raw scores since ratios are only appropriate for further analyses when a number of requirements are met (49). In an attempt to obtain more comparable results, the analyses were repeated on ˙VO2peak/weight, which resulted in lower heritability estimates. The relationship between˙VO2peak and weight, as well as other relationships between fitness and body measures and their underlying genetic and environmental causes, needs to be studied in more detail in a bivariate approach. Furthermore, different populations were considered with a different genetic pool living under different environmental conditions. The results reported by Devor and Crawford(15) were obtained from Ukranian immigrants in a Kansas Mennonite community where the use of alcohol and tobacco is prohibited. Such differences in life style may affect the relative contribution of transmissible and nontransmissible factors.
While our results on cardiorespiratory fitness agree with those of Fagard et al. (19) (heritability of 0.77), which are also based on twin data, Pérusse et al. (36,37) report much lower transmissibility estimates (0.22 and 0.28) for PWC150. The latter group found the transmissible variance to be entirely owing to cultural transmission. It should be noted that their data were corrected for body weight and that similar corrections of the data of Fagard et al.(19) and of our data reduced the reported heritabilities to 0.71 and 0.67, respectively. Other possible explanations for the better agreement between our results and those of Fagard et al.(19) than those of Pérusse et al.(37) are the use of an identical path model (ACE model)(22,33) on twin data versus a different path model(BETA model) (39) on twin and family data. Although Pérusse et al. (37) found substantial twin correlations, these data are modeled as a special twin environment instead of as a source of information for genetic variability. A few studies report husband-wife correlations on maximum oxygen uptake, but these are either nonsignificant (30) or negative(26), compared to the significant correlation of 0.42 in our sample.
Despite the differences in model assumptions between this study and other reports, Pérusse et al. (36,37) found somewhat higher and more comparable transmissibilities for push-ups (0.44), sit-ups (0.37-0.55), and trunk flexion (0.48), comprising both biological and cultural inheritance. Devor and Crawford (15) also reported fairly high transmissibility for trunk flexibiltiy (0.66). Pérusse et al. also report significant positive assortative mating for all fitness items studied which corresponds fairly well with our findings. Although the marital correlation of 0.10 for flexibility was significant in the study of Pérusse et al., the correlation of 0.09 between fathers and mothers in our study is not.
Adiposity has been studied more often by path-analytical model fitting to determine its genetic predisposition(3,4,11,12). The results of this study are compared with studies using skinfold measurements to determine adiposity. The fairly high heritability for the sum of skinfolds (0.87) is in contrast to the general tendency of results from path analysis to be lower than those of studies reporting heritability coefficients. One other exception is the study of Bodurtha et al. (3), who reported high estimates for the genetic variance of skinfolds similar to ours. Factors that may account for this similarity include the fact that in both the Leuven and the Virginia studies a sample of twins were born within a narrow age range and that the same path model was applied in both studies(22,33). Most other path-analytic studies use the TAU or BETA models (39), which allow the partitioning of the variance into transmissible and nontransmissible variance but not into genetic and environmental variance. The transmissibility includes both the heritability or biological inheritance and cultural inheritance. This model is applied to different family relationships such as parent-offspring pairs, sibling pairs (4), possibly including twin pairs(11,12). The samples of these studies are therefore less homogeneous for age since parents and children, and siblings, are mostly measured at different ages, which could result in lower heritability estimates. The inclusion of parents of twins in this study resulted in fairly high heritability estimates, but with evidence for different genes operating in 10-yr-old children and adults for subcutaneous fat.
Comparison of our findings with other twin and family studies involving fitness characteristics is difficult as diverse methods are used to estimate genetic variance, including traditional heritability estimates, intraclass correlations, or F-significance ratios of the genetic contribution. However, the maximum-likelihood correlations reported inTables 2 and 4 fall within the range of correlations reported in these twin(13,23-25,29,38,41,51) and family(7-9,26,27,30,35) studies. While the twin correlations are relatively high, the parent-offspring correlations are lower compared with previously published correlations. Possible explanations are the very narrow age range of the twins in this study and the large difference in age between the twins and their parents. Longitudinal twin data that are currently being collected should clarify these age-dependent genetic effects. Spouse correlations for motor characteristics in three Eastern European samples summarized by Malina(28) do not correspond well to those reported here. This may reflect differences in the selection of mates or differences in the effect of cohabitation in different populations. The results of this and other studies clearly indicate that for a number of fitness characteristics assortative mating is present and needs to be considered in the evaluation of genetic and environmental factors influencing motor performance.
The results should be interpreted in the context of three methodological limitations. First, although for most fitness variables our twin sample did not significantly differ from the Belgian population, it is unlikely that the sample is completely representative. The only adult reference data available are from a male sample of 30-yr-olds. While there may be a tendency for fit parents to participate in the study, it is probably not very strong since the overall fitness of the fathers was lower than that of the reference population. Second, the statistical power of the sample must be taken into account in interpretating the results. The fairly large sample of a specific age (10 yr) allowed us to reject a model that predicts that variation was solely a result of environmental factors. However, in the presence of substantial genetic variation, familial environmental factors, as well as genetic dominance, that account for a small proportion of the total variance would probably be undetectable. Adding parental data increases the power to detect shared environmental factors, as shown for the speed components. The twin-parent design also allows to take the effect of assortative mating into account. The large age difference between the twins and their parents can induce age-related variation, which was implemented as reduced genetic transmission. Longitudinal data on the twins will provide us with a more powerful design to model these effects.
Finally, fitness characteristics cannot be measured without error. To assess the influence of the measurement error, the total phenotypic variance is partitioned in reliable variance and unreliable variance, based on reliability coefficients obtained in a previous study(45). The reliable variance is then further partioned into genetic, common, and specific environmental variance. The heritabilities of the reliable phenotypic variance range from 0.56 for running speed (shuttle run) to 0.97 for static strength (arm pull). These results suggest that when fitness characteristics are measured without error the heritability estimates will be considerably higher.
In summary, this twin and family study supports the hypothesis that genetic factors explain the largest part of the variation in both performance-related and health-related fitness characteristics and that specific environmental factors account for the remaining variance for most items. However, the hypothesis that environmental influences are more important in explaining the variation in health-related than in performance-related fitness characteristics is not supported for this sample of 10-yr-old twins.
1. Baranowski, T., C. Bouchard, O. Bar-Or, et al. Assessment, prevalence, and cardiovascular benefits of physical activity and fitness in youth. Med. Sci. Sports Exerc.
2. Beunen, G., J. Borms, J. Vrijens, J. Lefevre, and A. L. Claessens. Physical Fitness and Sport Participation in Flemish Youths. Vol. 1: Physical Fitness of Flemish Youths 6 to 18 Years of Age
, Leuven, Belgium: Interuniversitair Onderzoekscentrum voor Sportbeleid, 1991, pp. 1-177.
3. Bodurtha, J. N., M. Mosteller, J. K. Hewitt, et al. Genetic analysis of anthropometric measures in 11-year old twins: The Medical College of Virgina twin study. Pediatr. Res.
4. Bouchard, C., A. Demirjian, and R. M. Malina. Path analysis of family resemblance in physique. Stud. Phys. Anthropol.
5. Bouchard, C. and R. M. Malina. Genetics of physiological fitness and motor performance. Exerc. Sport Sci. Rev.
6. Bouchard, C. and R. M. Malina. Genetics for the sport scientist: selected methodological considerations. Exerc. Sport Sci. Rev.
7. Bouchard, C., G. Lortie, J. A. Simoneau, C. Leblanc, G. Theriault, and A. Tremblay. Submaximal power output in adopted and biological siblings. Ann. Hum. Biol.
8. Bouchard, C. Genetic of aerobic power and capacity. In:Sport and Human Genetics
, R. M. Malina and C. Bouchard (Eds.). Champaign, IL: Human Kinetics, 1986, pp. 59-88.
9. Bouchard, C., R. Lesage, G. Lortie, et al. Aerobic performance in brothers, dizygotic and monozygotic twins. Med. Sci. Sports Exerc.
10. Bruce, R. A., J. R. Blackman, J. W. Jones, and G. Strait. Exercising testing in adult normal subjects and cardiac patients.Pediatrics
11. Byard, P. J., K. Sharma, J. M. Russell, and D. C. Rao. A family study of anthropometric traits in a Punjabi Community. II. An investigation of familial transmission. Am. J. Phys. Anthropol.
12. Clark, P., R. Jardine, N. G. Martin, A. E. Stark, and R. J. Walsh. Sex differences in the inheritance of some anthropometric characters in twins. Acta Genet. Med. Gemellol.
13. Crielaard, J. M. and F. Pirnay. Déterminisme génétique de l'aptitude physique. Trans. Soc. Franc. Méd. Sci. Sport
14. Derom, C., E. Bakker, R. Vlietinck, et al. Zygozity determination in newborn twins using DNA variants. J. Med. Genet.
15. Devor, E. J. and M. H. Crawford. Family resemblance for neuromuscular performance in a Kansas Mennonite Community. Am. J. Phys. Anthropol.
16. Eaves, L. J., A. C. Heath, N. G. Martin, et al. Comparing the biological and cultural inheritance of stature and conservatism in the kinships of monozygotic and dizygotic twins. In: Proceedings of 1994 APPA Conference
. C. R. Cloninger (Ed.), in press.
17. Eaves, L. J., K. A. Last, P. A. Young, and N. G. Martin. Model-fitting approaches to the analysis of human behaviour.Heredity
18. Engstrom, L. M. and S. Fischbein. Physical capacity in twins. Acta Gen. Med. Gemellol.
19. Fagard, R., E. Bielen, and A. Amery. Heritability of aerobic power and anaerobic energy generation during exercise. J. Appl. Physiol.
20. Fulker, D. W. Extensions of the classical twin method, In: Human Genetics, Part A: The Unfolding Genome (Progress in Clinical and Biological Research 103A)
, B. Bonne-Tamir (Ed.). New York: Alan R. Liss, 1982, pp. 395-406.
21. Garn, S. M., P. E. Cole, and S. M. Bailey. Living together as a factor in family-line resemblance. Hum. Biol.
22. Heath, A. C., M. C. Neale, J. K. Hewitt, L. J. Eaves, and D. W. Fulker. Testing structural equation models for twin data using Lisrel. Behav. Genet.
23. Klissouras, V. Heritability of adaptive variation.J. Appl. Physiol.
24. Komi, P. V. and J. Karlsson. Physical performance, skeletal muscle enzyme activities, and fibre types in monozygous and dizygous twins of both sexes. Acta Physiol. Scand.
25. Kovar, R. Genetic analysis of motor performance.Sports Med.
26. Lesage, R., J. A. Simoneau, J. Jobin, C. Leblanc, and C. Bouchard. Familial resemblance in maximal heartrate, blood lactate and aerobic power. Hum. Hered.
27. Lortie, G., C. Bouchard, C. Leblanc, et al. Familial similarity in aerobic power. Hum. Biol.
28. Malina, R. M. Genetics of motor development and performance. In: Sport and Human Genetics
, R. M. Malina and C. Bouchard (Eds.). Champaign, IL: Human Kinetics, 1986, pp. 23-58.
29. Malina, R. M. and W. H. Mueller. Genetic and environmental influences on the strength and motor performance of Philadelphia school children. Hum. Biol.
30. Montoye, H. J. and R. Gayle. Familial relationships in maximal oxygen uptake. Hum. Biol.
31. Neale, M. C. Mx Statistical Modeling. Box 3, 3 MVC
, Richmond, VA: Genetics and Human Development Technical Report, 1991.
32. Neale, M. C., A. C. Heath, J. K. Hewitt, L. J. Eaves, and D. W. Fulker. Fitting genetic models with Lisrel: hypothesis testing.Behav. Genet.
33. Neale, M. C. and L. R. Cardon. Methodology for Genetic Studies of Twins and Families
, Dordrecht, The Netherlands: Kluwer Academic Publishers, 1992, pp. 1-496.
34. Ostyn, M., J. Simons, G. Beunen, R. Renson, and D. Van Gerven. Somatic and Motor Development of Secondary Schoolboys: Norms and Standards
, Leuven, Belgium: Leuven University Press, 1980, pp. 45-48.
35. Pawlak, K. Body build heredity and sport achievements. In: Genetics of Psychomotor Traits in Man
, N. Wolanski and A. Siniarska (Eds.). Warsaw, Poland: College of Physical Education, 1984, pp. 111-123.
36. Perusse, L., C. Leblanc, and C. Bouchard. Inter-generation transmission of physical fitness in the Canadian population.Am. J. Sport Sci.
37. Perusse, L., G. Lortie, C. Leblanc, A. Tremblay, G. Theriault, and C. Bouchard. Genetic and environmental sources of variation in physical fitness. Ann. Hum. Biol.
38. Pirnay, F. and J. M. Crielaard. Influence de l'heredite sure les performances physiques. Med. Sport
39. Rice, J., C. R. Cloninger, and T. Reich. Multifactorial inheritance with cultural transmission and assortative mating: description and basic properties of unitary models. Am. J. Hum. Genet.
40. Sallis, J. F., T. L. Patterson, J. A. Morris, P. R. Nader, and M. J. Buono. Familial aggregation of aerobic power: the influence of age, physical activity and body mass index. Res. Q. Exerc. Sport
41. Schwarz V. Zwillingsuntersuchungen bei korperlichen Belastungen. Med. Sport
42. Simoneau, J. A., G. Lortie, C. Leblanc, and C. Bouchard. Anaerobic alactacid work capacity in adopted and biological siblings. In:Sport and Human Genetics
, R. M. Malina and C. Bouchard (Eds.). Champaign, IL: Human Kinetics, 1986, pp. 161-171.
43. Simons, J., G. Beunen, M. Ostyn, et al. Construction d'une batterie de test d'aptitude motrice pour garçons de 12 à 19 ans, par la méthode de l'analyse factorielle. Kinanthrop
. 1:323-362, 1969.
44. Simons, J., M. Ostyn, G. Beunen, R. Renson, and D. Van Gerven. Factor analytic study of the motor ability of Belgian girls age 12 to 19. In: Biomechanics of Sports and Kinanthropometry
, F. Landry and W. A. R. Orban (Eds.). Miami: Symposia Specialists, 1976, pp. 395-401.
45. Simons, J., G. P. Beunen, R. Renson, A. L. M. Claessens, B. Vanreusel, and J. A. V. Lefevre. Growth and Fitness of Flemish Girls: The Leuven Growth Study
. HKP Sport Science Monograph Series, Vol. 3,(Eds.). Champaign, IL: Human Kinetics, 1990, pp. 1-173.
46. Szopa, J. Familial studies on genetic determination of some manifestations of muscular strength in man. Genet. Polon.
47. Szopa, J. Inheritance of aerobic work capacity in man: results of population study on family resemblances. Genet. Polon.
48. Tanaka J. S. Multifaceted conceptions of fit in structural equation models. In: Testing Structural Equation Models
, K. A. Bollen and J. S. Long (Eds.). Newbury Park, CA: Sage Publications, pp. 10-39.
49. Tanner, J. M. The fallacy of per-weight and per-infance area standards and their relation to spurious correlation. J. Appl. Physiol.
50. Vlietinck, R. Determination of the Zygosity in Twins
, Ph.D. Thesis. Leuven, Belgium: K. U. Leuven, 1986.
51. Weiss, V. Die Heritabilitäten sportlicher tests, berechnet aus den Leistungen zehnjährigen Zwillingspaare.Leistungssport
52. Wolanski, N. Genetics and training possibility of psychomotor traits in man. In: Genetics of Psychomotor Traits in Man
, N. Wolanski and A. Siniarska (Eds.). Warsaw, Poland: College of Physical Education, 1984, pp. 21-52.
Keywords:©1996The American College of Sports Medicine
PERFORMANCE-RELATED FITNESS; HEALTH-RELATED FITNESS; STRENGTH, SPEED; BALANCE, FLEXIBILITY; CARDIORESPIRATORY FITNESS; ADIPOSITY; PATH ANALYSIS; Mx, MODEL-FITTING; HEREDITY; HUMAN VARIATION