Subjects were part of the Saskatchewan Bone Mineral Accrual Study (1991–97), which has been described in detail elsewhere (2). In brief, of the 375 eligible students attending two elementary schools in middle-class neighborhoods in Saskatoon, Saskatchewan, Canada, the parents of 228 children (113 boys and 115 girls) provided written consent for their child to be involved in this longitudinal study. The study utilized a mixed longitudinal design and incorporated eight age cohorts. The cohorts were aged between 8 and 15 yr at baseline. During the 7 yr of biannual or triannual data collection, the composition of these clusters remained the same. As there were overlaps in ages between the clusters, it was possible to estimate a consecutive 13-yr developmental pattern (8–21 yr) over the shorter period of 7 yr. Eligible children had no history of chronic disease or chronic medication use, and no medical conditions, allergies, or medications known to influence bone metabolism or calcium balance. The study received approval from the University and Hospital Advisory Committee on Ethics in Human Experimentation.
To be included in the analyses, each child required a valid determination of peak height velocity and complete measurements (biological age, stature, body mass, and physical activity) at more than one measurement occasion. Seventy boys and 68 girls fulfilled these requirements and comprise the data set for analyses. To ensure adequate cell sizes for each age band, only data between the ages of 9 and 18 yr were analyzed.
Chronological age (yr) was determined precisely to decimal age value. Decimal age was calculated by subtracting date of birth from the measurement date. Chronological age groups were constructed using 1-yr intervals such that the 10-yr age group included observations between 9.50 and 10.49 yr.
Anthropometric measurements, including height and body mass, were made every 6 months by International Society for the Advancement of Kinanthropometry (ISAK) certified personnel, according to the ISAK standards for anthropometric assessment (12). Subjects wore T-shirts and loose-fitting shorts during measurement, with shoes and jewelry removed. Height was measured as stretch stature using a wall stadiometer and recorded to the nearest 0.1 cm. Body mass was measured on a calibrated electronic scale and recorded to the nearest 0.1 kg. Two measurements were taken for each anthropometric variable. A third measurement was required if the first two differed by more than 0.4 cm for height and 0.4 kg for body mass. The two measurements for each anthropometric measure were averaged. If three measures were taken, the median value was used. After measurements were completed each individual’s results were compared with previously recorded results to ensure an increase or plateau in values.
Age at PHV is an indicator of somatic maturity and reflects the maximum growth in stature during adolescence. Measurement of height occurred in the spring and fall of each year (1991–97) with the intervening interval being 6 months (21) A growth curve was fitted to each individual’s annual height velocity data using a cubic spline procedure (GraphPad Prism version 3.00 for Windows, GraphPad Software, San Diego, CA) and age of PHV was determined (3). A cubic spline interpolates polynomials from information of neighboring points in an effort to obtain global smoothness. The cubic spline simply directs the curve through each data point. Although a mathematically fitted curve approach summarizes growth better and more elegantly (28), the approach used in this paper has the advantage of greater flexibility. Another advantage of the cubic spline technique is the integrity of the data is maintained such that each individual’s age of PHV is not modified or transformed by averaging group data. Longitudinal data were available in 70 boys and 68 girls.
A biological age was calculated by subtracting the chronological age at time of measurement from the chronological age at PHV. Thus, a continuous measure of biological age was generated. Biological age groups were constructed using 1-yr intervals such that the −1 PHV age group included observations between −1.49 and −0.50 yr from PHV. Physical activity measurements were considered in terms of years before and after PHV.
The Physical Activity Questionnaire for Older Children (PAQ-C) was administered a minimum of 3× yr−1 for the first 3 yr of the study and 2× yr−1 thereafter. The scores of the three or two assessments were averaged to create a single score to represent an individual’s level of physical activity for the year. The PAQ-C assesses general levels of physical activity and is described in detail elsewhere (3,6,16). In brief, the PAQ-C was designed for children in grades 3 or greater to assess their level of moderate and vigorous physical activity. Physical activity was defined as “sports, games, or dance that make you breathe hard, make your legs feel tired, or make you sweat” in the PAQ-C. In completing the PAQ-C, subjects were asked to rate their physical activity level during their spare time in the previous 7 d. Nine items scored on a 5-point Likert-type scale are averaged to derive an overall physical activity score ranging from one to five, with higher scores indicating higher levels of physical activity. The use of a 5-point rating in the PAQ-C results in a normal distribution of physical activity scores. For high school students, the PAQ-C was modified by omitting one item regarding physical activity at recess.
The PAQ-C had favorable 1-wk test-retest reliability in a sample of 84 children from grades 4 to 8 with intraclass correlation coefficients of r = 0.75 for boys and r = 0.82 for girls (6). Furthermore, the PAQ-C significantly correlated with other measures of physical activity in elementary (15) and secondary school students (16). Convergent validity for the PAQ-C was demonstrated in 89 students in grades 4 to 8, through a moderate relationship to a peer-comparison activity rating (r = 0.63), a 1-wk summation of 24-h moderate to vigorous physical activity recalls (r = 0.53), and perceptions of athletic competence (r = 0.48) (16). Further convergent and construct validity were demonstrated in the results of 97 elementary school students, ages 9–14 yr, through moderate relationships to a peer-comparison rating (r = 0.57), the Leisure Time Exercise Questionnaire (r = 0.41), a Caltrac motion sensor (r = 0.39), a 7-d activity recall interview (r = 0.46), and a step test of fitness (r = 0.28) (16). The Physical Activity Questionnaire for adolescents (PAQ-A) was also examined for convergent validity (15). In 85 students from grades 8 through 12, the scores from the PAQ-A were moderately related to a peer-comparison rating (r = 0.73), the Leisure Time Exercise Questionnaire (r = 0.57), a Caltrac motion sensor (r = 0.33), and the 7-d physical activity recall interview (r = 0.59).
Results are expressed as mean ± SD. A two-sided unpaired Student’s t-test was used for cross-sectional analysis with P < 0.05 considered statistically significant (SPSS version 10.0, SPSS Inc., Chicago, IL). A secondary longitudinal analysis, was performed using random effects linear modeling (MlwiN version 1.0, Multilevel Models Project, Institute of Education, University of London, London, UK) (9,10,17).
The following random effects regression model was adopted to describe the developmental changes in physical activity:EQUATION
where y is the PAQ-C summary score at assessment occasion i in the j th individual, αj is the constant for the j th individual, βjxij is the slope of PAQ-C with biological age for the j th individual, and k1 to kn are the coefficients of the various explanatory variables at assessment occasion i in the j th individual. εij is the level-1 residual (within individual variance) for the i th assessment of PAQ-C in the J th individual. Models were built in a stepwise procedure, that is, predictor variables (κ-fixed effects) were added one at a time. The difference in deviance of the likelihood ratio statistic between two models was compared with the probability of obtaining a chi square of this magnitude by chance alone to determine whether one model was a significant improvement over the other. In this way, predictor variables were added to the models and retained if deviance improved and/or if the variances at level 1 and level 2 were reduced (random effects in Tables 2 and 3). Predictor variables (κ) were accepted as significant if the estimated mean coefficient was greater than twice the standard error of the estimate (SEE), that is, P < 0.05. If the retention criteria were not met, the predictor variable was discarded. To allow for the nonlinearity of growth, age power functions were introduced into the linear models. Once growth and age were modeled, a sex effect was incorporated into the models as a categorical factor (boys = 0; girls = 1).
On average, each subject was observed on six annual occasions (median 6, range 4–7); in total, 850 physical activity observations were recorded. The distribution of the subjects’ physical characteristics represented at each chronological and biological age is reported in Table 1. From 9 to 13 yr of age, there were no significant differences between boys and girls in stature. However, from 14 yr and older, boys were significantly taller (P < 0.05). After 15 yr of age, boys had significantly more body mass (P < 0.05), with no significant differences observed before this age. When the data were assessed by biological age, boys were significantly taller and heavier at all ages pre- and post-PHV (P < 0.05).
When the physical activity data were analyzed in a cross-sectional fashion by chronological age bands, there were significant sex differences in PAQ-C summary scores from 10 to 16 yr of age (P < 0.05), with boys recording higher summary scores. In both sexes, PAQ-C summary scores decreased with increasing chronological age (Fig. 1a). When aligned by biological age, PAQ-C summary scores decreased with increasing biological age with no sex differences observed (P > 0.05) (Fig. 1b), except 3 yr before PHV (P < 0.05) when boys were more physically active.
Tables 2 and 3 summarize the results from the two multilevel models, one for each age categorization. The random effects model reports the total variance for the two levels of the model in PAQ-C summary scores (2), within individuals (level 1 of the hierarchy) and between individuals (level 2 of the hierarchy). In both models (Tables 2 and 3), the random effects were significant within individuals (i.e., estimated mean variance > 2*SEE), indicating that the PAQ-C summary scores were significantly decreasing at each measurement occasion within individuals (P < 0.05). The between individuals variance matrix (level-2 variance) for each model indicated that individuals had significantly different PAQ-C summary score curves in terms of their intercepts (constant/constant, P < 0.05) and the slopes of their lines (age/age, P < 0.05). The variance of these intercepts and slopes were not significant (constant/age, P > 0.05). The variance between individuals was therefore not different at different ages.
The coefficients that significantly predicted PAQ-C summary scores (fixed effects shown in Table 3) aligned by chronological age were chronological age and chronological age2; age powers were included in the models to shape the curves. In this model, height and body mass were not significant contributors of PAQ-C scores. A significant independent sex effect was found; girls had, on average, a 0.31 lower PAQ-C summary scores than boys (P > 0.05). When aligned on biological age (Table 3), biological age and biological age2 had independent effects on the prediction of PAQ-C summary scores. Again, height and body mass were not significant contributors of PAQ-C scores. In contrast to the chronological age model (Table 2) in the biological age model (Table 3), no sex difference was found (P > 0.05).
The results from this study confirm previous findings, including those from a longitudinal investigation of leisure time physical activity in an adolescent population (1), that physical activity levels in children decreased with increasing age from ages 9 to 18 yr. When examined using chronological age bands, girls’ physical activity levels were lower than boys’. However, when aligned on a biological age bands (years from PHV), sex differences were no longer apparent, apart from 3 yr before PHV. The results from this cross-sectional analysis at separate biological age time points were confirmed when the data were analyzed using a random effects modeling approach.
There is large variation in the timing and tempo of maturation in boys and girls. Regardless of early or late maturation, the sequence of maturity events is closely adhered. Biological age can be assessed by various methods including sexual, morphological, dental, and skeletal criteria. Although no single system of biological maturation provides a complete description of the maturation of an individual child, interrelationships between systems are, nevertheless, strong enough to indicate the biological development level of a group of children (19). As in most other longitudinal growth studies (5,13,26,27) subjects were aligned along an axis related to age of attainment of PHV. Age at PHV is a very useful maturity indicator serving as a landmark against which attained sizes and velocities of other body composites can be expressed (20). The relationship between growth in height and the development of reproductive organs is much stronger than the relationship of either with growth of pubic hair (23,27). In relation to skeletal age, a higher correlation between chronological age at PHV and skeletal age at PHV was noted in girls (r = 0.74) than in boys (r = 0.34) (19). Because individuals advanced or delayed in sexual maturation are also advanced or delayed in linear growth, it is recognized that age from PHV is a useful indicator of biological age (27). Other than the measures of skeletal or dental development, age of PHV is the only maturity assessment that can be used to make direct comparisons between boys and girls. In the present sample, peak height velocity was achieved, on average, 1.7 yr earlier in females than in males. These findings are consistent with other studies (20,28). Although the girls’ average age at PHV (11.8 yr) was similar to previously reported values from U.S. and European populations, the boys in the present study reached PHV (13.4 yr) roughly 5–6 months earlier than previously reported (20).
The finding that level of physical activity decreased with increasing chronological age in boys and girls was in accordance with the existent literature. A unique feature of the results presented is that they offer further insights into the development of physical activity with maturity and body size appropriately controlled. For a comprehensive understanding of the changes in physical activity with growth, changes in physical activity should ideally investigate the influences of growth covariates simultaneously. In the random effects model when individuals were aligned by chronological age (Table 2), neither body mass nor stature, independent predictors of physical activity in the growing child, were significant. However, physical activity was decreasing with increasing chronological age, as indicated by the negative, statistically significant, age coefficient (−0.13 [SEE 0.03] total physical activity score). When the confounders of body size and age were controlled, girls were found to have, on average, 0.31 (SEE 0.09) total physical activity scores less than boys at all ages. Although the model confirms previous analyses that boys (marginal mean = 3.11 PAQ-C score) are significantly more active than girls (marginal mean = 2.71 PAQ-C score) (6), the biological significance of a 0.31 PAQ-C score has yet to be investigated. However, in a children’s physical activity intervention study (grades 4, 5, and 6), it was found that PAQ-C scores significantly increased by a 0.31 PAQ-C score in the treatment compared with the control group (8). This suggests that a 0.31 difference in PAQ-C score does indeed reflect a difference in physical activity levels. When aligned on biological age (Table 3 model b), body mass and stature were nonsignificant, independent predictors of physical activity, whereas biological age was significantly and negatively associated with physical activity. This finding demonstrated that physical activity also decreased with increasing biological age. In contrast to the chronological age model (Table 2), no significant, independent differences were found between boys and girls in PAQ-C scores when aligned on biological age (Table 3).
Research examining the level of physical activity in children is limited by the difficulties associated with the various assessment techniques (4). The measurement of physical activity itself is difficult because it occurs in a variety of forms and contexts. Measures of physical activity such as motion detectors and heart rate monitors provide an objective assessment of level of physical activity. Limitations of these techniques include mechanical failure, children tampering with the devices, reactivity, and cost. Self-report may be the most practical and cost effective method, but no standard exists. The major limitation of self-report questionnaires in assessing physical activity is the accuracy of recall. Younger children particularly may be influence by their ability to recall specific physical activities complete with estimates of intensity and duration.
The PAQ-C has been rigorously tested for reliability and validity (6,7,15,16). Presently, there is no single method identified as the gold standard for measuring physical activity in children and adolescents. As such, the PAQ-C was considered as appropriate as other methods available for providing an indication of an overall level of physical activity in children ages 8 and older. The authors are confident that the patterns of physical activity observed in the current investigation were real and not a reflection of measurement error associated with the questionnaire.
The results of the current investigation provide further evidence that the level of physical activity declines with age in both males and females from 8 to 18 yr of age. In agreement with other studies, at similar chronological ages, males were more active than females. However, this gender difference in level of physical activity disappeared when maturational differences between the sexes were taken into consideration. This is a new finding that calls for further study.
In this study, the comparison of boys and girls physical activity levels while controlling for biological age illuminated the influence of pubertal development on physical activity behaviors. In both boys and girls, and at the same rate, physical activity decreased with increasing biological age. Although the results of the present investigation cannot provide specific answers as to why this occurred, there are myriad cultural, behavioral, and physiological factors operating during adolescence that may be related to this. The lack of opportunity, availability, and cultural support for nontraditional physical activities may partly explain the trend noted. Clearly, the association between physical activity, behavior and maturation is an important area that warrants further investigation.
This work was supported by National Health and Research Development Program (NHRDP) grant no. 6608-1261, Canada.
1. Aaron, D. J., K. L. Storti, R. J. Robertson, A. M. Kriska, and R. E. Laporte. Longitudinal study of the number and choice of leisure-time physical activities from mid to late adolescence: implications for school curricula and community recreation programs. Arch. Pediatr. Adolesc. Med. 156: 1075–1080, 2002.
2. Bailey, D. A. The Saskatchewan Pediatric Bone Mineral Accrual Study: bone mineral acquisition during the growing years. Int. J. Sports Med. 18: S191–S194, 1997.
3. Bailey, D. A., H. A. Mckay, R. L. Mirwald, P. R. E. Crocker, and R. A. Faulkner. A six-year longitudinal study of the relationship of physical activity to bone mineral accrual in growing children: the University of Saskatchewan Bone Mineral Accrual Study. J. Bone Miner. Res. 14: 1672–1679, 1999.
4. Baranowski, T., and B. G. Simons-Morton. Dietary and physical activity assessment in school-aged children: measurement issues. J. School Health 61: 195–197, 1991.
5. Boas, F. Studies in growth. Hum. Biol. 4: 307–350, 1932.
6. Crocker, P. R. E., D. A. Bailey, R. A. Faulkner, K. C. Kowalski, and R. Mcgrath. Measuring general levels of physical activity: preliminary evidence for the Physical Activity Questionnaire for Older Children. Med. Sci. Sports Exerc. 29: 1344–1349, 1997.
7. Crocker, P. R. E., and R. A. Faulkner. Self-report of physical activity intensity in youth: gender and grade level considerations. Avante 5: 43–51, 1999.
8. Ernst, M. P., and R. P. Pangrazi. Effects of a physical activity program on children’s activity levels and attraction to physical activity. Pediatr. Exerc. Sci. 11: 393–405, 1999.
9. Goldstein, H. Multilevel Statistical Model, 2nd Ed. London: Edward Arnold, 1995, pp. 1–178.
10. Goldstein, H., J. Rasbash, I. Plewis, et al. A User’s Guide to MlwiN
. London: Multilevel Models Project, Institute of Education, University of London, 1998, pp. 1–281.
11. Hovell, M. F., J. F. Sallis, B. Kolody, and T. L. Mckenzie. Children’s physical activity choices: a developmental analysis of gender, intensity levels, and time. Pediatr. Exerc. Sci. 11: 158–168, 1999.
12. International Society for the Advancement in Kinanthropometry (ISAK). International Standards for Anthropometric Assessment. Underdale SA, Australia: ISAK, 2001, pp. 1–133.
13. Kemper, H. C. G. Literature survey of longitudinal growth research. Med. Sport Sci. 20: 1–11, 1985.
14. Kimm, S. Y. S., N. W. Glynn, A. M. Kriska, et al. Longitudinal changes in physical activity in a biracial cohort during adolescence. Med. Sci. Sports Exerc. 32: 1445–1453, 2000.
15. Kowalski, K. C., P. R. E. Crocker, and N. P. Kowalski. Convergent validity of the physical activity questionnaire for adolescents. Pediatr. Exerc. Sci. 9: 342–352, 1997.
16. Kowalski, K. C., P. R. E. Crocker, and R. A. Faulkner. Validation of the physical activity questionnaire for older children. Pediatr. Exerc. Sci. 9: 174–186, 1997.
17. Kreft, I., and J. de Leeuw. Introducing Multilevel Modeling. London: Sage, 1998, pp. 1–149.
18. Lindquist, C. H., K. D. Reynolds, and M. I. Goran. Sociocultural determinants of physical activity among children. Prev. Med. 29: 305–312, 1999.
19. Malina, R. M. Adolescent growth and maturation: selected aspects of current research. Yearbook Phys. Anthropol. 21: 63–94, 1978.
20. Malina, R. M., and C. Bouchard. Growth, Maturation, and Physical Activity. Champaign, IL: Human Kinetics, 1991, pp. 1–501.
21. Mirwald, R. L., and D. A. Bailey. Seasonal height velocity variation in boys and girls 8–18 years. Am. J. Hum. Biol. 9: 709–715, 1997.
22. Mota, J., and P. Queiros. Children’s behavior: physical activity regarding parents’ perception vs. children’s activity. Int. Rev. Soc. Sport 31: 173–178, 1996.
23. Reynolds, E. L., and J. V. Wines. Physical changes associated with adolescence in boys. Am. J. Dis. Child 82: 529–547, 1951.
24. Riddoch, C. J., J. M. Savage, N. Murphy, G. W. Cran, and C. A. G. Boreham. Long term health implications of fitness and physical activity patterns. Arch. Dis. Child 66: 1426–1433, 1991.
25. Sallis, J. F., J. E. Alcaraz, T. L. Mckenzie, and L. F. Hovell. Predictors of change in children’s physical activity over 20 months. Am. J. Prev. Med. 16: 222–229, 1999.
26. Stolz, H. R., and L. M. Stolz. Somatic Development of Adolescent Boys. New York: Macmillan, 1951, pp. 1–257.
27. T anner, J. M. Growth at Adolescence.
Oxford: Blackwell Scientific, 1962, pp. 1–325.
28. T anner, J. M. Foetus into Man: Physical Growth from Conception to Maturity.
London: Castlemead, 1989, pp. 1–250.
29. Trost, S. G., R. Pate, J. F. Sallis, et al. Age and gender differences in objectively measured physical activity in youth. Med. Sci. Sports Exerc. 33: 350–354, 2001.
30. Van Mechelen, W., and H. C. G. Kemper. Habitual physical activity in longitudinal perspective. In: The Amsterdam Growth Study, H. C. G. Kemper (Ed.). Champaign, IL: Human Kinetics, 1995, pp. 135–158.