A twin study on the correlates of voluntary exercise behavior in adolescence

Objectives To improve the success of interventions aimed to increase moderate to vigorous physical activity, we need to better understand the correlates of the extensive individual differences in voluntary exercise activities. Starting in adolescence, genetic effects become a dominant factor in explaining individual differences in voluntary exercise behavior. Here we aim to establish the prospective contribution of potential correlates of voluntary exercise behavior to its heritability. Design In a sample of adolescent and young adult twins, data on potential correlates of exercise behavior were collected using surveys (time point 1, N = 373) and a laboratory study (time point 2, N = 499). Information on personality, perceived barriers & benefits, subjective and objective exercise ability and the affective response to exercise were collected in a set of healthy adolescent twin pairs (16-18y) and their non-twin siblings (12-25y). Almost 3 years later, the subjects were sent an online follow-up survey on their current exercise status (time point 3, N = 423). Methods In bivariate models, the phenotypic (co)variance in these correlates and exercise behavior at all time points were decomposed in sources of genetic (co)variance and environmental (co)variance. The correlates that were significantly associated with exercise behavior at time point 1 or 2 and showed significant genetic correlations to exercise behavior at time point 3 were used in two further analyses: Multiple regression analysis to predict exercise behavior at time point 3, and a genetic analysis in a common 2-factor model, that tested the overlap in genetic factors influencing these correlates and exercise behavior. Results Personality (Extraversion), perceived benefits and barriers, exercise-induced affective response (Energy measured after the cycling test), and subjective and objective exercise ability (VO2max) showed significant phenotypic and genetic association with exercise behavior at time point 3. The genetic correlation between the two latent factors in the common 2-factor model was 0.51, indicating that part of the heritability in exercise behavior derives from genetic variants that also influence these correlates. Conclusions Given their shared genetic basis and predictive power we assert that individual differences in extraversion, perceived benefits and barriers, exercise-induced feelings of energy, and subjective and objective exercise ability can be used to develop stratified interventions for adolescent and young adult exercise behavior. In addition, our results provide the first clues on ‘where to look’ for specific genetic variants for voluntary exercise behavior.