The determinants of physical activity level are not well understood. Demographic factors such as age, education, and socioeconomic status are all related to physical activity level ⁽²³⁾. Social and physical environment, psychology, and physiology also play a role⁽²²⁾. Familial factors, both genetic and environmental, may play a role in determining the level of physical activity engaged in by an individual. Other risk factors for chronic disease, notably alcohol consumption and body mass index, are known to have sizable genetic components^(14,16,24). Such findings are neither meant to preclude the preception that these traits are modifiable nor to suggest that interventions are inappropriate. Common environment is also thought to play a role in risk factors such as diet and smoking^(12,14,19). Few studies have investigated whether voluntary physical activity, that not associated with employment, has significant familial components.

In this study, we use the Vietnam Era Twin Registry to address and quantify familial clustering of voluntary physical activity in men age 33 to 51; and we explore partitioning familial aggregation into genetic and common environmental components. Both moderate physical activity and intense forms of exercise are investigated to determine whether genetics and environment contribute to physical activity level in middle age.

SUBJECTS AND METHODS

Subjects. Study subjects are twins who are members of the Vietnam Era Twin Registry. These are men who both served in the U.S. military between May, 1965 and August, 1975, either in Southeast Asia or elsewhere. At the time that these data were collected in 1990 subjects' ages ranged from 33 to 51. The registry includes a total of 7,350 male-male twin pairs. The VET Registry was developed to investigate the long-term effects of military service in Vietnam. A complete description of the Vietnam Era Twin Registry can be found elsewhere ⁽¹¹⁾.

The first survey of the VET Registry in 1987 used a mixed-mode mail and telephone survey to collect data on physical and psychological health. In total, the 1987 survey achieved a individual response rate of 74%. In 1990 a second mail and telephone survey identified prevalent cases of cardiovascular disease and baseline risk factor data for future studies. To be included in this study, both members of the pair must have completed the 1987 survey. This 1990 cardiovascular survey achieved an individual response rate of 75% (3344 twin pairs). The purpose of each study was described fully to survey participants.

At the time of the survey, the mean age of the subjects was 41 yr (range, 33 to 51), and the mean educational attainment was two years of post-secondary school. Seventy-eight percent of the participants were married, and 95% were white. The median 1987 household income for participants was between $30,000 and $34,999: 19% had an income above $50,000, and 4% had a household income below $10,000.

Zygosity was assigned based on a series of questions on twin similarity and limited blood group typing obtained from military records. The use of self-reported similarity questions to assign zygosity has been demonstrated to be a valid method when compared with blood typing. The details of the VET Registry methodology are fully discussed elsewhere⁽⁸⁾.

Physical activity measurements. Physical activity was assessed by two sets of questions. The first set of six questions addressed whether there was consistent effort to increase physical activity level through choosing more active options in the course of the day, focusing on walking and stair climbing. These questions are in the spirit of the 1995 guidelines from the National Centers for Disease Control and Prevention and the American College of Sports Medicine, recommending moderate physical activities such as“walking up the stairs instead of taking the elevator, walking instead of driving short distances, doing calisthenics, or pedaling a stationary bicycle while watching television” ⁽¹⁸⁾.

We assessed moderate activity with the following questions:

Do you usually participate in any of the following activities?

• Climbing stairs instead of taking the elevator

• Walking instead of driving short distances

• Parking away from your destination so you have to walk more

• Walking on your lunch break or after dinner

• Getting off at a bus stop before your destination and walking

• Other extra walking or stair climbing for exercise.

The second set of five questions asked about regular performance of intense forms of exercise for specified durations, exercises which are considered vigorous and require 4.5 or more metabolic equivalents⁽¹⁷⁾. These questions relate to the 1978 guidelines from the Department of Health and Human Services and the American College of Sports Medicine which recommended a physical activity level defined on the basis of intense exercise, enough to ensure fitness, as assessed by maximum oxygen uptake ⁽³⁾.

We assessed intense activity with the following questions:

For at least the last three ⁽³⁾ months, which of the following activities have you performed regularly?

• Jog or run at least 10 miles per week

• Play strenuous racquet sports (singles tennis, paddle ball, etc.) at least 5 hours per week

• Play other strenuous sports (basketball, soccer, etc.)

• Ride a bicycle at least 50 miles per week

• Swim at least 2 miles per week.

To assess whether groups of questions could be combined into a composite scale, we calculated Cronbach's coefficient alpha separately for the moderate and intense activity questions ⁽⁶⁾. The alpha coefficient, a measure of reliability based on correlations, indicates whether there is internal consistency for the items included in a composite scale. In this study the alpha coefficient for the summed index of the six moderate activity questions was 0.62, while the coefficient for the summed index of the five intense activity questions was 0.32. A zero to six scale was constructed for the number of positive responses to the moderate activity questions; however, no single scale was constructed for the intense activity questions, as the low alpha coefficient suggests that the individual question items are not measuring the same underlying phenotype.

Analytic strategy. The analysis strategy is directed to answering two of the fundamental questions in genetic epidemiology⁽¹⁵⁾: 1) Does a trait cluster in families? and 2) Is familial clustering related to common environmental exposure, biologically inherited susceptibility, or cultural inheritance?

In twin studies the extent of familial aggregation can be assessed by the within-pair similarity for a phenotype, such as a specific leisure-time physical activity. The degree to which both twins tend to be concordant for a trait is an indication of familial clustering.

To further explore the sources of familial resemblance, the special characteristics of twinship are exploited. Twin pairs are the same age and share the same birth family environment (unless adopted or reared separately), known as common or shared environment. In addition, monozygotic twins are identical for all genetic factors, while dizygotic twins like siblings share, on average, half their genes. Classic twin studies assume that the common environment has the same effect for MZ and DZ^(13,15) (an assumption addressed below). By comparing how similar members of MZ pairs are to how similar members of DZ pairs are with respect to the phenotypic trait, one can partition the variability in the population into three inferred sources: genes, common environment, and environment unique to each person.

Since MZ pairs have identical genes and common environment, the extent to which they differ is the estimator for unique environmental effects for both MZ and DZ pairs ⁽²¹⁾. If there is no correlation between either MZ or DZ twin pairs, then unique environment alone determines observed phenotype. When there is a positive correlation between twins, and the correlation is the same for MZ pairs and DZ pairs, then one infers that the familial effect is wholly a result of common environment rather than genes. When the MZ correlation is greater than the DZ, that is taken as evidence of a genetic effect. If the familial aggregation results from genes alone, then the MZ correlation will be at least twice the DZ correlation. When the genetic effect is the result of additive influences of alleles at a number of genetic loci, the MZ correlation will be exactly twice the DZ correlation. If, however, alleles at genetic loci interact, nonadditive or dominance genetic variation results. Since a dominance genetic effect correlates 0.25 in DZ twins but perfectly in MZ twins, its presence will reduce the DZ correlation to be less than one-half the MZ correlation. One assumes that common environment does not account for any of the observed similarity when the MZ correlation is greater than twice the DZ correlation (although it is also possible that this pattern is the result of common environment playing a greater role in MZ than DZ pairs).

Bearing these ratios in mind, one can simultaneously solve equations to calculate heritability (the proportion of the phenotypic variance in the population attributable to genotypic or allelic differences among individuals)⁽¹⁵⁾ and common environment in terms of the MZ and DZ correlations. Heritability in the broad sense is the sum of additive and nonadditive genetic effects.

To compensate for potential deviations from the assumption that the common environment is comparable in MZ and DZ pairs, we restrict the estimation of heritability to pairs who see each other regularly as adults, at least once per month (1,006 MZ pairs and 530 DZ pairs). This approach was taken to reduce the likelihood of overestimating heritability as a result of greater adult contact and concomitant lifestyle similarities for MZ than DZ pairs. We recognize that the single measure of adult contact may not completely capture the differences between MZ and DZ pairs; however, more detailed measures were unavailable.

Statistical methods. Initial descriptive statistics present the percent of individuals who participated in each of the specific moderate and intense activities separately for MZ and DZ pairs. Next, a familial aggregation odds ratio is calculated, expressing the relative odds of a twin engaging in a specific leisure-time physical activity when his co-twin also engages in this same activity. The maximum likelihood estimator of the odds ratio obtained from a logistic regression analysis modified for twin studies is used to represent the extent of familial aggregation. Ninety-five percent confidence intervals for the familial odds ratios that do not include unity indicate the significance of familial clustering⁽²¹⁾.

To explore the relative contribution of genetic and environmental effects, we calculate polychoric twin correlations between MZ and DZ pairs for each of the physical activity measures by maximum likelihood methods⁽¹³⁾. These correlations are used to estimate heritability and common environmental effects.

Hypothesis testing for genetic and common environmental effects is carried out using logistic regression analysis adapted for genetic analysis of twin data ⁽⁷⁾. Adjustment by age and race had no effect on the results and is not reported. All analysis was performed using SAS System 6.07(SAS Institute, Cary, NC).

RESULTS

For each of the moderate activity questions, the percent who answered positively is given in Figure 1. Percentages vary little by zygosity. For all but one of the moderate activity questions the percent responding positive is high, ranging from 30.3% for walking on your lunch break in dizygotic twins to 66.4% for climbing stairs instead of taking the elevator in monozygotic twins. Getting off at a bus stop before your destination and walking, which could only be answered positively by men who regularly take buses, was positive for just 3.5% of DZ and 4.3% of MZ respondents. The questions inquiring about intense, regular exercise had lower positive participation rates (Fig. 2). Swimming at least 2 miles·wk^-1 had the lowest percent of participants (3.2% in DZ and 3.3% in MZ), while the heterogeneous, other strenuous sports had the highest positive participation (12.8% in MZ and 13.8% in DZ).

The odds ratios for each of the physical activity questions are statistically significant, indicating familial clustering(Table 1). For the moderate activities, odds ratios consistently range from 1.4 to 1.9, with the exception of the question about getting off the bus before the destination and walking, which has an odds ratio of 4.3. The odds ratio for the index of moderate activities is also statistically significant. The odds ratios are higher for the intense exercises, ranging from 3.0 for the heterogeneous strenuous sports question to 4.6 for biking.

For each of the moderate and intense activities, correlations were higher for the MZ than the DZ pairs, suggesting genetic influence(Table 2). MZ correlations ranged from 0.27 for other extra walking or stair climbing for exercise to 0.58 for racquet sports. For DZ pairs, the lowest correlation was 0.07 for jogging or running, while the highest, 0.44, was for getting off the bus before the destination. Heritability was significant for three of the moderate activity questions, for the moderate activity scale, and for the moderate activity scale, and for one of the intense activity questions, jogging or running; estimates of heritability in the broad sense range from 27% for other extra walking or stair climbing to 53% for jogging or running. While high heritabilities are derived for the intense activities such as riding a bicycle or racquet sports, these are not significant, perhaps because of low statistical power.

DISCUSSION

Despite mounting scientific evidence and well-publicized public health recommendations, few adults exercise regularly. The purpose of this study is to determine whether there is familial clustering of physical activity level and what roles heredity and common environment, which includes the effect of the early years twins spend together, play in voluntary physical activity among middle-aged men.

Two types of physical activities are studied. The first type represents moderate physical activity consciously undertaken to increase activity level, such as choosing to climb stairs rather than take an elevator or choosing to walk rather than drive short distances. The second type represents intense exercise, such as running 10 miles·wk^-1 or swimming 2 miles·wk^-1. A significant limitation of this study is that the physical activity questions have not been validated in this population although validation studies have been carried out with energy expenditures computed from survey instruments including similar questions (1,2.25).

A further limitation of this study is the special nature of the study population. All were male veterans who were healthy and fit when they entered the military. They are more active than the general population; 24.6% of the men answered yes to at least one of the five intense exercise questions. Too few nonwhites were included in the Registry to detect any differences by race; just 4.5% of the study group was black. The age range represented is narrow so that age was not associated with physical activity. Since both members of a twin pair had to respond to two surveys to be included here, there is also the possibility of nonresponse bias. However, a comparison of 1990 survey responders with nonresponders finds the same mean age (41 yr), education (14 yr), and household income ($30,000-$34,999). However, respondents were more likely to be married (78% vs 66%) and more likely to be white (95% vs 87%).

This study provides evidence of a strong familial effect for specific forms of intense exercise and of a consistent but modest effect for moderate physical activities. To partition the familial effect into genetic and environmental components, MZ and DZ twin pairs with regular adult contact are contrasted. Restricting the analysis to these pairs does not preclude the possibility that heritability estimates are inflated because MZ pairs share a more similar environment as adults than DZ pairs even though both sets of pairs see each other at least once a month. Statistical power, however, is a problem in quantifying genetic effects for activities with low prevalence in the study population.

For all of the measures of moderate and intense physical activities, MZ correlations are higher than DZ correlations, which suggests that heredity plays a role in adult physical activity level. Common environmental effects were not observed in the present study although one must bear in mind that some of the variation attributed to genetic effects may in fact reflect more similar MZ lifestyles as adults. When responses to each of the physical activity questions are analyzed separately, genetic effects are generally 12% stronger for the intense exercises than for the moderate ones. For running or jogging, racquet sports, and bicycling, heritability is between 48% and 58%, although only the 53% for running or jogging is significant. For six measures of moderate physical activity, heritability estimates ranged from 12% to 40%; the index of moderate activity is 38% heritable. In these data, common environment is not significantly associated with the phenotypic variability in the moderate or intense physical activities. Common environment includes the nongenetic effects which MZ and DZ twins share equally, such as social class, education, and parental behavior. If MZ pairs have more similar environments as adults or as children than DZ pairs, common environment is underestimated in these models. Also, given the standard errors of the underlying twin correlations, the estimates of heritablity and common environment are suggestive and not definitive.

Few studies have examined genetic and environmental influences on physical activity. Previous results are not strictly comparable as physical activity is assessed through different instruments. A Finnish study of 1,537 MZ and 3,507 DZ male twin pairs measured physical activity by asking for the individual's opinion of the amount, intensity, and duration of current physical activity and the number of years of activity in adult life. A factor analysis was performed and a single factor constructed. Intraclass correlations for the factor were 0.57 for MZ twins and 0.26 for DZ twins, yielding a heritability estimate of 0.62 and a common environmental effect of zero⁽¹²⁾.

The Finnish twin study contrasts both in method and findings with that of Pérusse et al. ⁽²⁰⁾ who collected 3-d activity records of 1,610 blood relatives and adoptees from 375 Quebecois families between 1978 and 1981. For each 15-min period of the 3 d (including one weekend day), subjects recorded the energy expenditure of the primary activity on a scale from one to nine using a list of categorized activities. From these records, two activity scores were computed: the mean daily sum of the 15-min scores and the mean number of daily periods with an activity score from 6 to 9, indicating moderate to intense work or sports activities⁽²⁰⁾. For level of habitual physical activity, the first phenotype, heritability, accounted for 29% of the variance in the model; for exercise and sports activity, the second phenotype, there was no genetic effect and common environment accounted for 12% of the variance.

The two phenotypes of Pérusse et al. ⁽²⁰⁾ are not unrelated to the two types of activities assessed in this study, and the results are compatible. Their heritability estimate for level of habitual physical activity, 29%, is not inconsistent with the 38% found here for moderate activity level. With respect to the construction of a single additive measure of intense activities, their second phenotype, the low value for Cronbach's alpha coefficient for intense exercises in the present study, suggests the possibility that different forms of intense exercise may not all reflect the same phenotype. Thus, a combined measure would yield a low heritability estimate.

Three small studies have examined genetic and environmental determinants of sports participation. An Australian study of 200 twin pairs assessed genetic influences on several lifestyle risk factors, including a single exercise question, “vigorous exercise in past 2 weeks”⁽¹⁰⁾. Heritability was estimated at 0.39 for this question although the authors were concerned that MZ pairs had greater adult contact than DZ pairs, a violation of the equal environment assumption in twin studies. Using the single question “Have you been involved in sports activities during the last three months?”, Boomsma et al.⁽⁵⁾ analyzed responses from 90 adolescent Dutch twin pairs and their parents. For females, heritability was 35%; for males, heritability was 77%. A single model estimated heritability at 64%. The remainder was explained by unique environment. Gedda determined twin concordance rates among 128 same sex male-male and female-female twin pairs, both for any sports participation and for participation in the same sport⁽⁹⁾. While he does not present heritability, by calculating tetrachoric correlations from his data, heritability can be estimated as zero for any sports participation, but 60% for same sports participation, similar to the findings here for particular intense activities. The inconsistencies in these studies may reflect both small sample sizes and dissimilar ascertainment of sports participation.

The present study suggests that a portion of the phenotypic variability seen in the population with respect to deliberately maintaining a moderately active lifestyle, consistent with new guidelines, may be genetically mediated. It is beyond the scope of our data and this study to suggest by what mechanism or mechanisms physical activity could be influenced by genes and whether the effect is mediated through physical, physiologic, or psychological traits. While we found evidence for some genetic influence on moderate and intense activities, the majority of the variability in our data is not associated with genetic effects, but rather with individual environmental factors. These data suggest the possibility that regular participation in specific sports such as running, racquet, and other strenuous sports may be influenced by genetics more than moderate activities, although sample size is a problem for sports with low participation rates.