The promotion of lifelong physical activity (PA) and healthy lifestyle is among the main aims of physical education in many countries (25). Investment in physical education is based on the belief that PA in youth is beneficial for young peoples’ health, and also becomes habitual and tracks over time, and thus influences individual and public health in the adult population. Many transitions and life-changing events during the life course affect PA, and therefore, the level of tracking of PA is likely to vary in different phases of life. Thus, information about the tracking of PA may be useful in planning interventions in different phases of life. Tracking is usually defined as a tendency of individuals to maintain their rank or position within a group over time (18). Tracking also means the ability to predict subsequent values on the basis of earlier values (8). Here, the term stability is used as parallel with tracking.
Interest in the tracking of PA has increased notably during the past decade. Most tracking studies have been published since the year 2000 (30). Despite the increase in tracking research generally, little additional information about the long-term stability of PA has been reported as the follow-up time in most studies has been short (30). Even there are many studies in which the follow-up time has been more than 20 yr, there is, to our knowledge, no study where PA has been followed from the age of 3 yr to adulthood (older than 27 yr). Also, the many measurements between the baseline and the final follow-up allowing to detect the level of tracking in different phases of life are rare in previous studies.
Previous results have shown that, among men, PA has significant but low or moderate stability during single life phases, and also during longer periods of life. In women, the level of tracking has been lower and in many studies nonsignificant. PA stability seems to be lower in early childhood than that in adolescence or in adulthood and lower in transitional phases, such as from childhood to adolescence or from adolescence to adulthood, than that in adulthood (30).
In most tracking studies, self-report methods have been applied, and the results have not usually been adjusted for measurement errors or other error variance. Therefore, the real stability of PA may be greater than indicated by unadjusted self-report results. This is supported by the higher correlations found in studies where PA has been measured using objective methods (15,22,31).
The aim of this study was to investigate the tracking of PA from preschool age to adulthood in six age cohorts. To our knowledge, this is the first study which analyze PA tracking in both genders and with many age cohorts during the 27-yr follow-up period, and which covers tracking from preschool age to adulthood.
MATERIALS AND METHODS
Data were obtained from the Young Finns Study that was launched in 1980 and was repeated in 1983, 1986, 1989, 1992, 2001, and 2007 (26). Altogether, 4320 children and adolescents age 3, 6, 9, 12, 15, and 18 yr were randomly chosen from the population register of five university cities and their surroundings to produce a representative sample of Finnish children. Of these subjects, 3596 (83%) participated in the initial survey and were followed until 2007. To keep the follow-up intervals as equal as possible for the simplex modeling (see Statistical analysis section), only five measurements, 1980, 1986, 1992, 2001, and 2007, were selected for the present study (Table 1). The study protocol was reviewed and approved by the ethics committee of each of the five participating univesities and the subject’s informed conset was obtained (26).
The PA of 3- and 6-yr-old children was measured using mothers’ ratings. Mothers were asked questions concerning their child’s outdoor play time (h·d−1) in summer and in winter, the amount of PA in play as compared with other children, the vigorousness of PA, the child’s enjoyment of indoor/outdoor play, the child’s general level of activity as compared with other children, the encouragement given to participate in sports, and the patterns of PA. Each item was coded from 1 to 3, except for encouragement to engage in sport (1–2). By summing the variables, a PA index (PAI) of preschool children was formed with scores ranging from 8 to 23.
The PA of 9- to 18-yr-old subjects in 1980 and 1986 was measured using a short self-report questionnaire administered individually in connection with a medical examination. The questions concerned the frequency and intensity of leisure-time PA, participation in sports club training, participation in sport competitions, and habitual way of spending leisure time. The items were coded from 1 to 3 and summed to form a PAI with scores ranging from 5 to 15 (28). In 1992, 2001, and 2007, the PA questionnaire consisted of items on the frequency of PA, intensity of PA, frequency of vigorous PA, hours spent on vigorous PA, average duration of a PA session, and participation in organized PA (29,33). The PAI was calculated in the same way as for the youth groups. Internal consistency coefficients (α) as indicators of reliability varied from 0.44 to 0.76 at baseline, the coefficients being lower in the younger cohorts than that in the older ones. In 2007, the coefficients varied from 0.72 to 0.82. Reliability coefficients were also calculated by the simplex model (Table 3). The validity of the PAI was tested by showing statistically significant correlation with the indicators of exercise capacity in a subsample (n = 102) (29) and an inverse correlation with waist circumference in both genders (35). The convergent validity of the PAI in 2007 was also shown by its correlation with pedometer step counts. The correlations were not high in men (r = 0.26, P < 0.001) and women (r = 0.32, P < 0.001) (13), but considering that the PAI was an overall estimate of the leisure-time and commuting activities, the pedometer counts were only a 1-wk sample of activity, and the pedometer did not react to many usual Finnish activities such as biking, swimming, skiing, fitness club, and so on, the correlations would not be expected to be very high. In addition, the predictive validity of the PAI in relation to health factors has been published previously (32,35). There were no significant difference in the baseline PA between participants and dropouts in 2001 (26,29).
Spearman rank-order correlations (Spearman rho) between the PAI measured in subsequent years were calculated and tested for significance. A simplex model was used to test the hypothesis that PA patterns remain stable over time (i.e., PA of each measurement time depends on the PA of previous measurement time; Fig. 1). Simplex model was fitted separately for males and females in each cohort using self-rated PA for time interval from 1980 to 2007 in four oldest cohorts (Fig. 1A) and from 1986 to 2007 in 3- and 6-yr-old-cohorts (Fig. 1B).
Simplex modeling enables also measurement error (et) to be distinguished from real change. It is possible to divide the variance of the observed variables (PAIt) into measurement error variance and the variance of latent (real) variables (pait) at the time of measurement (t). The stability coefficients (stabt) reflect the extent to which pait could be predicted from pait−1, whereas the term ζt is a residual term (16). This means that it is also possible to evaluate the reliability of each measurement and the correlations between the latent variables, which could then be interpreted as the coefficients of the stability of the real variables.
Moreover, the indirect stability coefficient for the 27-yr interval was calculated in four oldest cohorts using simplex model. The indirect stability coefficient is the correlation between the latent variables in the first and the last occasion. The corresponding Spearman rank-order correlation for the 27-yr interval was calculated and tested for significance.
The indirect effects of latent PA in 1980 on latent PA in 2007 mediated through years 1986, 1992, and 2001 were estimated and tested using simplex models in the four oldest cohorts. The indirect effect or the mediated effect between the first and the last occasion would be equal to the product of the direct effects.
Because the measure of PA among 3- and 6-yr-olds in 1980 was different from other years, the indirect effect in two youngest cohorts was calculated so that latent PA in 1986 was predicted by mother-rated PA in 1980, and from 1986, the model was applied as in other cohorts. The indirect effect of mother-rated PA on the latent PA in 2007 through the measurements was estimated and tested for significance.
To evaluate the goodness of fit of the models, the χ2 test, the comparative fit index (CFI), the Tucker–Lewis index (TLI), the standardized root mean square residual (SRMR), and the root mean square error of approximation (RMSEA) were used. A model fitted the data well when the P value associated with the χ2 test was nonsignificant. CFI and TLI values close to 0.95, SRMR values lower than 0.08, and RMSEA values lower than 0.06 indicated a good fit between the hypothesized model and the observed data (14).
In addition, a more parsimonious model, in which the direct effects (unstandardized regression coefficients between adjacent latent variables) were constrained to be equal, was estimated. The χ2-difference test was conducted for the nested models. If the restricted model was not rejected in favor of the less-restricted model, the estimation results of the more parsimonious model were reported. To end up with an identifiable model, the measurement error variances of the observed PAI in the first and second occasion were estimated as equal; similarly, the measurement error variances of the observed PAI in last two occasions were estimated as equal. Furthermore, when the model gave negative or nonsignificant residual variances of the latent PAI or measurement error variances, these parameters were fixed to be close to zero (0.01).
The analyses were performed by using the IBM Statistical Package for the Social Sciences version 19.0 (IBM, Armonk, NY) and the Mplus statistical package version 6.1 (Muthén & Muthén, Los Angeles, CA). The standard MAR approach (missing at random) using the full-information maximum likelihood estimation was applied to simplex modeling (21).
Mean and SD for the PAI are presented in Table 2. The PAI for ages 3 and 6 yr are different because of the differences between mother report and self-report. Also, the PAI in 1992, 2001, and 2007 differed little from those in 1980 and 1986. In the cohorts of 9- to 18-yr-olds, the PAI declined clearly from 1980 to 1986, particularly in males. These changes ware small from 1992 to 2001 and for females even upward, but the PAI declined and remained relatively low from 2001 to 2007.
The more parsimonious simplex model with the equal regression coefficients was confirmed in every cohort of boys and girls. The simplex model seemed to fit the data reasonably well. The χ2-test was nonsignificant in all age–sex groups, except for the cohort of 9-yr-old boys and 15-yr-old girls. CFI was higher than 0.95 for all models and TLI was 0.94 or higher, except for the cohort of 15-yr-old girls. SRMR ranged from 0.03 to 0.09 and RMSEA from 0 to 0.07 for all models (data not shown).
The reliability coefficients of all the measurements calculated by the simplex model are presented in Table 3. The coefficients of 3- and 6-yr-old kids in 1980 were omitted because of the use of different PA scale. The coefficients of reliability were very high in the two youngest age–sex groups with one exception in 1986. These coefficients probably were overestimated because of measurement error variances fixed to be close to zero. The “real” or true values for the age–sex groups would be close to the more realistic estimates of reliability in other birth cohorts at the same age (0.41–0.64).
The coefficients of stability in which measurement error had been taken into account were higher than the traditional Spearman rank-order correlations within the 6-yr interval for all age groups in both sexes (Table 4). Among males, there was a steadily increasing trend in the stability with age. The stability coefficients were consistently 0.60 or higher after the age of 18 yr in each cohort of males. Among females, there was similar but not as clear increasing trend in the stability with age. From 2001 to 2007, the stability coefficients were very high in the cohorts of 9-, 12-, and 15-yr-old males and 18-yr-old females. The overestimated coefficient was partly due to the fact that the residual variances in 2007 were fixed to be near zero. However, the Spearman correlations between the PAI were significant in all age–sex groups. In particular, the correlations were remarkably well with its highest level in most of the age–sex groups within the last 6-yr interval ranging from 0.49 to 0.61 in males and from 0.36 to 0.50 in females.
The indirect stability coefficient of PA between 1980 and 2007 ranged from 0.28 to 0.44 in males and from 0.08 to 0.45 in females. Among boys, for instance, at age 18 yr, the latent variable explained 9% of the variance of the latent variable at age 45 yr. Among 18-yr-old girls, the corresponding proportion was 20%. The coefficients of stability were also higher than the Spearman correlations within the 27-yr interval for all age groups in both genders, except the 9-yr-old boys and girls.
Mother-rated PA in 1980 predicted latent PA in 1986 in 3- and 6-yr-old boys (unstandardized regression coefficients b = 0.20, P < 0.001, R2 = 0.08 and b = 0.24, P < 0.001, R2 = 0.11, respectively) and in 3-yr-old girls (unstandardized regression coefficient b = 0.13, P < 0.001, R2 = 0.05), but not in 6-yr-old girls (b = 0.07, P = 0.12, R2 = 0.01). Moreover, mother-rated PA of 3- and 6-yr-old boys in 1980 indirectly predicted the PA in adulthood (in 2007) via the PA in childhood, adolescence, and young adulthood (Table 5). In four oldest cohorts, unstandardized estimates of the indirect effects of latent PA between the first and the last occasion ranged from 0.25 to 0.43 in males and from 0.12 to 0.45 in females. The highest estimate was observed for 15-yr-old boys and 12-yr-old girls. The indirect effects of latent PA were significant in all age groups of males and in 12- and 18-yr-old females.
To study the stability of activity (high activity) versus inactivity (low activity), the subjects (four older cohorts) were divided into tertiles according to the PAI in 1980, 1992, and 2007. The percentages of those remaining in the same tertile from 1980 to 2007 were slightly higher in the activity than the inactivity groups in males (46.7% vs 41.4%) and females (41.5% vs 36.9%). During the period 1980–1992, the activity group was more stable than the inactivity group in both males (45.4% vs 32.9%) and females (45.6% vs 32.6%). Later in adulthood, from 1992 to 2007, no stability difference was observed between the activity and the inactivity groups in females, whereas in males, the inactivity group showed slightly higher stability (62.2%) than the activity group (51.9%) (data not shown).
This study investigated how PA tracks from early childhood to adulthood in six birth cohorts of men and women. To our knowledge, this is the first PA tracking study to follow several cohorts starting from preschool age and continuing for 27 yr. The main result was that the 6-yr integrated PA stability coefficients adjusted for measurement errors were moderate or high in youth and high in adulthood. The long-term direct and indirect effects of PA in 1980 on the PA in 2007 were significant but low in all cohorts of males and in two cohorts of females. The indirect effect of mother-rated PA of the 3- and 6-yr-old children on the adult PA 27 yr later was significant but weak in males but not in females.
In many studies where the subjects have been followed for more than 20 yr from a young age to adulthood, the tracking correlations have been nonsignificant or very low (3,10,12,19,24). In the present study, the coefficients of stability from 1980 to 2007 (0.25–0.45), in which the measurement error has been taken into account, were higher than previously reported tracking correlations, except the value of 0.08 found in the birth cohort of 9-yr-old girls. The tracking correlations in adulthood (18–45 yr) were similar (0.14–0.19) to those of previous studies on long-term tracking in adults (10,17,24). Our stability coefficients in adulthood (0.30–0.45) resemble those found in a recent study with a similar follow-up time (20). The variation in tracking correlations between countries may mean that there are cross-cultural differences in the tracking of PA. It seems that in Nordic countries where also the level of PA is rather high also the stability of PA is higher than that in many other countries.
The tracking of PA during and from preschool age has been studied using both objective methods and ratings by mothers. The latter have produced both significant (11) and nonsignificant results (23) across a short interval. In our study, the tracking correlations for mother-rated PA of 3-yr-olds were 0.52 for boys and 0.61 for girls, with subsequent measurements at 3-yr intervals, indicating good test–retest reliability for the mothers’ ratings (data not shown). This is in the line with a recent review (15), which reported moderate to high tracking correlations of objectively measured PA among very young children. The findings that mother-rated PA at age 3 yr significantly predicted self-rated PA at age 15 yr in boys and at age 24 yr in girls (data not shown), show rather good predictive validity for the mothers’ ratings. This was supported by the fact that the indirect effect between mother-reported PA in 1980 and PA 2007 was significant in 3-yr-old and 6-yr-old boys’ cohorts. This suggests that the habitual pattern of PA starts to develop very early during preschool age and also that mothers are aware of their children’s PA and are capable to evaluate it. Maternal observations for birth weight, age of walking, and other simple indicators of children’s development are often routinely used in studies. Our results show that mothers are able to report much more complicated information about their children and their daily life, for example, children’s PA in indoor and outdoor play and in other daily activities. Mothers’ insight to children’s activities could perhaps be better used for research.
Earlier studies have reported rather little variation by age in tracking correlations or stability coefficients (20,30). However, in this study, the interage correlations and stability coefficients from age 9 to 12 yr are lower than those from age 18 to 21 yr, indicating that PA is less stable in childhood and early adolescence than that in adulthood (as shown in Table 4), as has also been found to a certain extent in some earlier studies (30). The age differences in the stability coefficients are smaller than the differences in the correlations because the stability coefficients have been corrected for reliability, which is lower in the younger groups than older groups.
The gender effect on the tracking of PA was apparent in all phases of the life course, and also in the long term, males showed greater stability than females. This has also been found in some previous studies (4,31), but not consistently. A recent study found a significant 10-yr tracking among girls but not in boys (9). It is difficult to see any clear reason for the gender difference in the tracking of PA. In general, gender difference in the level of PA is small or nonexistent in Finland (5). It is possible that some kind of gender inequity exists in the capacity of females to maintain their PA level during major transitions and life changes. This is supported by findings that many life changes, such as getting married or having small children in the family, have a greater influence on the PA of women than that of men (1,34). Some recent studies have found that among young people, the gender differences are smaller if the biological age instead of chronological age is taken into account, and also the differences in the tracking of PA are smaller (7,9).
PA has usually been tracked through coefficients, for example, ranking correlations, showing to what extent individuals maintain the same position in the activity distribution. From the point of view of physical education and public health, an important question is how far active people stay active and inactive people stay inactive. As stated by Corbin (6), more attention should be paid to the tracking of inactivity. Our results indicated that activity tracks slightly better than inactivity in the younger phases of life (1980–1992), unlike the results of previous studies where inactivity seems to track better than activity (2,19,27). However, in our study, the stability of inactivity was slightly higher than the stability of activity in males in later adulthood (1992–2007), which is in line with previous results. The difference between the tracking of activity and that of inactivity was small in our study, and it has also been small in previous studies. To draw valid conclusions on this issue, more research is called for on the stability of high PA and of low activity or inactivity at different life phases.
An important question related to tracking of activity or inactivity is what in general causes tracking and what kinds of determinants regulate tracking. There is very little research based information about this issue. Our results show that the level of tracking is related to age and gender. In particular, the stability of youth was lower than that of adulthood, which may be related to the decreasing level of PA, indicating that some people decrease more than others. One explanation for this may be that young people who participate regularly in organized sport may keep a part of their physically active longer than those who do not. However, it is important to emphasize that the tracking and changes of PA level are different issues. In our data, the low stability was related to the chance of activity in youth but in adulthood the relationship was different. From 2001 to 2007, the level of PA decreased clearly but the stability coefficients were high. It is also noteworthy that tracking results of PA are reported especially when accompanied with its changes.
Although objective instruments for measuring PA are being increasingly used, most long-term PA tracking studies are based on self-reports. As objective measures of PA become more widespread, it will be possible to study the tracking of PA in a more reliable and valid way. At present, the comparison of the results of objective measurements and self-reports allows us to believe that PA is more stable than is indicated by the unadjusted results from self-report studies. Measurement errors in self-report studies cause lower tracking correlations and thus give lower stability values than is in fact the case. The use of simplex models is one way to take measurement error into account and estimate of the “true” stability of PA. The simplex model, and our unique design with several measurements, also made it possible to calculate the indirect effect of base line PA through the measurements 1986, 1992, and 2001 on the PA in 2007, which, in case of 3- and 6-yr-old boys, gave a really important result showing the connection between the preschool age PA and the adult PA in males.
The strengths of this study are the long follow-up time, 27 yr, a representative sample covering ages from 3 to 45 yr, six birth cohorts, and the use of several measurements, allowing us to use simplex model to analyze the stability of PA at different phases of life course and long-term indirect effects. A limitation is that self-ratings or mothers’ ratings have been used to measure PA.
In conclusion, this study has shown that physically active lifestyle starts to develop very early in childhood and that the stability of PA is moderate or high along the life course from youth to adulthood.
This study was financially supported by the Academy of Finland (grants no. 77841, 210283, 121584, and 124282), the Social Insurance Institution of Finland, the Ministry of Education and Culture, the Turku University Foundation, the Special Federal Grants for Turku University Hospital, the Juho Vainio Foundation, the Finnish Foundation of Cardiovascular Research, the Finnish Medical Foundation, the Finnish Cultural Foundation, and the Yrjö Jahnsson Foundation.
The authors declare that they have no financial conflict of interests to disclose with any of the companies or manufacturers named in this article.
The results of the present study do not constitute endorsement of any product by the authors or the American College of Sports Medicine.
1. Allender S, Hutchinson L, Foster C. Life-change events and participation in physical activity: a systematic review. Health Promot Int
. 2008; 23: 160–72.
2. Anderssen N, Wold B, Torsheim T. Tracking of physical activity in adolescence
. Res Q Exerc Sport
. 2005; 76: 119–29.
3. Beunen GP, Lefevre J, Philippaerts RM, et al. Adolescent correlates of adult physical activity: a 26-year follow-up. Med Sci Sports Exerc
. 2004; 36 (11): 1930–6.
4. Boreham C, Robson PJ, Gallagher AM, Cran GW, Savage JM, Murray LJ. Tracking of physical activity, fitness, body composition and diet from adolescence
to young adulthood
: the Young Hearts Project, Northern Ireland. Int J Behav Nutr Phys Act
. 2004; 1: 14.
5. Borodulin K, Laatikainen T, Juolevi A, Jousilahti P. Thirty-year trends of physical activity in relation to age, calendar time and birth cohort in Finnish adults. Eur J Public Health
. 2008; 18: 339–44.
6. Corbin CB. The “untracking” of sedentary living: a call for action. Pediatr Exerc Sci
. 2001; 13: 347–56.
7. Erlandson MC, Sherar LB, Mosewich AD, Kowalski KC, Bailey DA, Baxter-Jones ADG. Does controlling for biological maturity improve physical activity tracking? Med Sci Sports Exerc
. 2011; 43 (5): 800–7.
8. Foulkes MA, Davis CE. An index of tracking for longitudinal data. Biometrics
. 1981; 37 (3): 439–46.
9. Francis SL, Morrissey JL, Letuchy EM, Levy SM, Janz KF. Ten-year objective physical activity tracking: Iowa bone development study. Med Sci Sports Exerc
. 2013; 45 (8): 1508–14.
10. Friedman HS, Martin LR, Tucker JS, Criqui MH, Kern ML, Reynolds CA. Stability of physical activity across the lifespan. J Health Psych
. 2008; 13: 1092–104.
11. Hallal PC, Wells JCK, Reichert FF, Anselmi L, Victora CG. Early determinants of physical activity in adolescence
: prospective birth cohort study. BMJ
. 2006; 332: 1002–7.
12. Herman KM, Craig CL, Gauvin L, Katzmarzyk PT. Tracking of obesity and physical activity from childhood
: the physical activity longitudinal study. Int J Pediatr Obes
. 2009; 4: 281–8.
13. Hirvensalo M, Telama R, Schmidt MD, et al. Daily steps among Finnish adults: variation by age, sex, and socioeconomic position. Scand J Public Health
. 2011; 39: 669–77.
14. Hu L, Bentler PM. Cutoff criteria for fit indexes in covariance structure analysis: conventional criteria versus new alternatives. Struct Equ Modeling
. 1999; 6: 1–55.
15. Jones RA, Hinkley T, Okely AD, Salmon J. Tracking physical activity and sedentary behavior in childhood
: a systematic review. Am J Prev Med
. 2013; 44: 651–8.
16. Jöreskog K, Sörbom D. LISREL 8 User’s Reference Guide
. Chicago: Scientific International, Inc. 1999. pp. 277–96.
17. Kirjonen J, Telama R, Luukkonen R, Kääriä S, Kaila-Kangas L, Leino-Arjas P. Stability and prediction of physical activity in 5-, 10-, and 28-year follow-up studies among industrial employees. Scand J Med Sci Sports
. 2006; 16: 201–8.
18. Malina RM. Physical activity and fitness: pathways from childhood
. Am J Hum Biol
. 2001; 13: 162–72.
19. Matton L, Thomis M, Wijndaele K, et al. Tracking of physical fitness and physical activity from youth to adulthood
in females. Med Sci Sports Exerc
. 2006; 38 (6): 1114–20.
20. Morseth B, Jørgensen L, Emaus N, Jacobsen BK, Wilsgaard T. Tracking of leisure time physical activity during 28 yr in adults: the Tromsø study. Med Sci Sports Exerc
. 2011; 43 (7): 1229–34.
21. Muthen B, Muthen L. Mplus User’s Guide and Mplus Version 6.11 (1998–2010)
. 2010. pp. 197–232. Available from: http://www.stamodel.com
22. Nyberg G, Ekelund U, Marcus C. Physical activity in children measured by accelerometry: stability over time. Scand J Med Sci Sports
. 2009; 19: 30–5.
23. Oja L, Turimäe T. Tracking of motor abilities, physical actaivity, and elementary motor skills during transition from pre-school to school. Acta Kinesiol Univ Tartuensis
. 2001; 6: 91–101.
24. Parsons TJ, Power C, Manor O. Longitudinal physical activity and diet patterns in the 1958 British Birth Cohort. Med Sci Sports Exerc
. 2006; 38 (3): 547–54.
25. Puhse U, Gerber M, (Eds). International Comparison of Physical Education: Concepts, Problems, Prospects
. Aachen: Meyer & Meyer Verlag. 2005. pp. 19–31.
26. Raitakari OT, Juonala M, Rönnemaa T, et al. Cohort profile: the cardiovascular risk in Young Finns Study. Int J Epidemiol
. 2008; 37: 1220–6.
27. Raustorp A, Svenson K, Perlinger T. Tracking of pedometer-determined physical activity: a 5-year follow-up study of adolescents in Sweden. Pediatr Exerc Sci
. 2007; 19: 228–38.
28. Telama R, Yang X, Laakso L, Viikari J. Physical activity in childhood
as predictor of physical activity in young adulthood
. Am J Prev Med
. 1997; 13: 317–23.
29. Telama R, Yang X, Viikari J, Välimäki I, Wanne O, Raitakari O. Physical activity from childhood
: a 21-year tracking study. Am J Prev Med
. 2005; 28: 267–73.
30. Telama R. Tracking of physical activity from childhood
: a review. Obes Facts
. 2009; 2: 187–95.
31. Tudor-Locke C, Giles-Corti B, Knuiman M, McCormack G. Tracking of pedometer-determined physical activity in adults who relocate: results from RESIDE. Int J Behav Nutr Phys Act
. 2008; 5: 39.
32. Yang X, Telama R, Hirvensalo M, Mattsson N, Viikari JS, Raitakari OT. The longitudinal effects of physical activity history on metabolic syndrome. Med Sci Sports Exerc
. 2008; 40 (8): 1424–31.
33. Yang X, Telama R, Hirvensalo M, et al. Leadership component of type A behavior predicts physical activity in early midlife. Int J Behav Med
. 2012; 19: 48–55.
34. Yang X, Telama R, Leino M, Viikari J. Factors explaining the physical activity of young adults: the importance of early socialization. Scand J Med Sci Sports
. 1999; 9: 120–7.
35. Yang X, Telama R, Viikari J, Raitakari OT. Risk of obesity in relation to physical activity tracking from youth to adulthood
. Med Sci Sports Exerc
. 2006; 38 (5): 919–25.