Significant changes in physical activity (PA) and aerobic power (V̇O2max) occur as youth mature from 8 through 16 yr of age. Cross-sectional studies have documented that PA levels generally decline at rates ranging from 1% per year to over 20% per year (1,12,27). Bradley et al. (5) have reported that as youth aged from 8 to 16 yr, the proportion of youth engaged in no vigorous PA almost doubled from 12 to 22%. In addition, the number of sedentary activities increased, such that by the high school years, talking on a telephone and television viewing rank high in the activities most reported. Conversely, not all studies agree that PA levels decline, as Janz et al. (10) and Ross et al. (22) report small increases (4–9%) in PA levels in males and females from 10 to 15 yr of age.
Aerobic power (V̇O2max) also appears to decline as youth age. Cross-sectional data suggest that although absolute V̇O2max (mL·min−1) increases as the individual increases in size, the V̇O2max relative to body mass (mL·kg−1·min−1) declines (19,26). McMurray et al. (19) reported a 34% and 53% increase in absolute V̇O2max for females and males, respectively, as they age from 8 to 16 yr, but V̇O2max relative to body mass declined 22% for females and 11% for males during the same time interval.
Most of the data on PA levels and aerobic power of youth were obtained using cross-sectional samples (1,26,27). Although many of these studies provided sufficient sample size to make general conclusions, specific year-to-year longitudinal changes are difficult to determine. In addition, some cross-sectional samples were obtained during the same year in different age groups, or using age groups that were separated by a number of years, e.g., 12 yr olds versus 16 yr olds (1,6), or fifth and sixth graders versus eighth and ninth graders (3,27). Thus, there appear to be little data from investigators who have followed a group of youth as they age from elementary school through high school. Such longitudinal data could give a better impression as to the possibility of a critical time point whereupon exercise declines, which could have implications for focusing future interventions.
A longitudinal data set allows investigators to track individuals, to see how their exercise habits change over time, and to determine the tendency of individuals to maintain their rank within a group over time (14). To this end, Kimm et al. (12) recruited 2379 females at 9–10 yr of age and followed them for 10 yr. They noted a 35% decline in PA levels as indicated by a 3-d recall and an 83% decline in PA as noted by a survey of habitual PA. However, no effort was made to compute tracking coefficients. The Young Finns Study (28) has shown that the correlation between PA levels (as measured by questionnaire) when youth were 9–12 yr old and when they were 12–15 yr old ranged from 0.4 to 0.55. The Amsterdam Growth and Health Longitudinal Study (11) followed subjects from 13 to 33 yr of age and found that PA tracked poorly, with a stability coefficient of 0.35. With respect to relative aerobic power measurements, Kemper et al. (11) reported a significant beta coefficient for aerobic power when they compared the V̇O2max at 13 yr of age with those obtained at 33 yr of age. The Leuven Growth Study (15,16) reported high correlations for V̇O2max (r = 0.78 to 0.86) as youth aged from 12 to 18 yr, and good correlations as the youth aged to 30 yr (r ∼ 0.63).
The reason for the better tracking of V̇O2max compared with PA appears to be the genetic component of aerobic power compared with the large behavioral component of PA levels. Also, PA is usually obtained via questionnaires, which normally have more variability. Although these studies indicate moderate tracking of V̇O2max does occur, there appear to be large breaks, or gaps, in the timing of data gathering and the majority of information is from adolescents and not from younger aged youth. In addition, ethnic differences in tracking have not been evaluated, yet studies have shown that ethnic differences in aerobic power and physical activity levels exist (8,19,30). Therefore, the purpose of this study was to evaluate the tracking of physical activity levels and aerobic power in African-American and Caucasian youth as they age from 8 to 16 yr.
METHODS
Subjects.
The subjects for this study were participants from the Cardiovascular Health in Children Study: CHIC (9). CHIC is a longitudinal study that evaluated the development and trajectories of cardiovascular disease risk factors in youth and adolescence. Parental consent and youth assent, using forms previously approved by the University’s Internal Review Board, were obtained before participation in the study. The initial data were obtained on over 2100 youth in third and fourth grade (age range 7–11 yr) from 21 elementary schools across the state of North Carolina during 1990. The initial sample was approximately half rural and half urban, containing a racial mix consistent with the demographics of North Carolina (19.4% African-American, 76.3% Caucasian, 4.3% other). From this initial sample, 529 females and 535 boys (N = 1064) were available in which follow-up data were obtained a minimum of three times over the next 7 yr, or to approximately 16 yr of age. These youth were used to assess the year-to-year change in V̇O2max and PA levels. In addition, subset of these youth who were in the study at years 1 and 7 (387 girls and 404 boys; N = 791) were used for the overall analyses (14).
The most common reason for withdrawing from the study was moving to a different location. Table 1 shows some of the characteristics of the subset of subjects that completed the initial and final testing in comparison with those that were lost to follow-up. In general, the body mass indexes (BMI), V̇O2max, and PA scores were similar (t-test:P > 0.05). The proportions of youth of low SES (parents with less than high school education) from the subjects that remained and those lost to follow-up were similar for the girls, regardless of ethnicity. In contrast, more low SES African-American boys remained in the study than were lost (χ2:P < 0.05).
TABLE 1: Comparison of the mean (±SD) body mass index (BMI), aerobic power, physical activity scores, as well as the proportion of sample with low SESa obtained the subset of youth that completed the testing (N = 1061) with those that were lost at follow-up (N = 1146).
Instrumentation.
Physical activity levels were obtained from a physical activity survey, adapted over the years for the reading comprehension level of the students. Youth in grades 6–10 completed a survey that listed 32 different activities youth commonly perform and how often they completed each activity (7). Activity scores were developed by multiplying how often the activity was performed by the MET level of the activity as described by Ainsworth et al. (2) in the Compendium of Physical Activity. We understood that adult MET levels underestimate energy expenditure for youth and adolescents; however, the Compendium was the most comprehensive list available. In addition, any error would be systematic and not change the subject-to-subject, or year-to-year variability. In 8- to 10-yr-old youth (grades 3–5), PA data were obtained using another instrument; thus, we did not include PA information from this age group in any analyses. PA scores ranged from 20 to 689, with the higher score representing a higher level of PA. Oxygen uptake (mL·kg−1·min−1) was predicted from a multi-stage cycle ergometer test using methods we have previously reported (18,19). Stature (±0.1 cm) and body mass (±0.1 kg) were measured on the youth at their respective schools, using a stadiometer and a calibrated balance beam scale, respectively.
Procedures.
The data were obtained throughout the school year by a team of trained research assistants (RA) who went into the schools to collect the information. Seasonality was accounted for as much as possible by scheduling the visit to each specific school during the same time of year (fall, winter, or spring) for each of the six measurement periods. During each measurement period, the students first reported in small groups (10–20 youth), and completed the physical activity recall and other questionnaires under the supervision of an RA. The students were removed from their classes at various times throughout the day to have height and weight measured, and to complete the exercise test. Other variables, unrelated to this study, were also obtained at this time. For a complete description, see Harrell et al. (9).
Analyses.
Means and standard deviations (mean ± SD) were computed for each variable for each year of testing. ANOVA for gender and ethnic differences in relative aerobic power (V̇O2max; mL·kg−1·min−1) as well as PA score were first computed for descriptive purposes. Mixed model ANOVA analyses were also computed within each gender and ethnic grouping to determine trends with respect to time (repeated measures for six times and independent for gender or ethnicity). When the ANOVA analyses for time, time by gender interaction, or time by race interaction were significant (P < 0.05), a post hoc Newman-Keuls test was employed to determine related differences.
Analyses were computed to determine the year-to-year degree of tracking for PA and aerobic power. Aerobic power included year 1 through 7, whereas PA used year 2 through 7. The loss of the first year’s data for PA was related to using a different instrument for assessing PA. First, year-to-year Spearman rank order correlations were computed within genders and ethnicities. According to Maia et al. (15), these Spearman correlations (ρ) represent stability estimates for tracking. These analyses included all youth for whom there were data in the two successive trials. Because V̇O2max and PA levels change as youth age, the subjects were divided into year-by-year gender-specific tertiles (low, moderate, and high categories) for aerobic power, as well as PA. Kappa (κ) statistics were computed to determine the likelihood that a particular youth would be classified in the same group from year to year (20). For edification, Munoz and Bangdiwala (20) have shown that κ values greater than 0.75 would be considered “almost perfect” agreement, 0.45–0.74 would be considered “substantial” agreement, scores of 0.20–0.44 are considered moderate agreement, and <0.20 “fair or poor” agreement. From the subset of youth that completed all six trials we computed: 1) the Spearman correlations (ρ) between the first and last years, 2) the Shrout-Fleiss reliability intraclass correlation (25) between the first and last years, 3) the κ statistic for stability between the first and last years, and 4) the stability coefficients derived from the β-estimate of the generalized estimating equations (GEE) as used by the Amsterdam Growth and Health Study (11,29). Twisk et al. (29) have shown that the stability coefficient obtained by the GEE analyses can be interpreted as a correlation coefficient ranging from 0 (no relationship) to 1 (perfect relationship). Thus, the closer the coefficient is to 1.0, the stronger the tracking. Because the yearly correlations between V̇O2max and PA were low (r∼0.15–0.20), the GEE analyses did not use either V̇O2max or PA score as covariates for each other. Using the subset that completed first and last testing, we also computed the percentages of youth that remained in their initial classification (year 1 for V̇O2max and year 2 for PA) at the end of the study (year 7). These analyses were completed by gender and ethnicity. All analyses were computed using SAS (SAS, Cary, NC).
RESULTS
The physical characteristics of the subjects are presented in Table 2 by gender. In general, the girls and boys had similar BMIs; however, the girls were shorter and weighed less than the boys. The girls also had lower physical activity scores than the boys and lower aerobic power. With regard to ethnicity, the African-American girls weighed more and had greater BMI than the Caucasians. There was a tendency for the V̇O2max of the African-American girls to be lower than the Caucasians (P = 0.064), particularly for years 4 through 7. No differences in V̇O2max were evident for the boys. In general, the African-American girls were also less active than the Caucasians (P = 0.03), but the African-American boys were more active than the Caucasians (P = 0.028). Although the ANOVA for PA levels were significant for both genders, post hoc means comparisons between ethnicities were not different for any of the six measurements.
TABLE 2: Physical characteristics and activity levels of the subjects at each of the six trials presented by sex and ethnicity.
The year-to-year Spearman correlations (ρ) for V̇O2max and physical activity level were all significant for both sexes and ethnicities; however, the correlations were higher for V̇O2max than physical activity (see Table 3). The correlations for V̇O2max between years 1 and 7 were similar for boys and girls, regardless of ethnicity. The year 2-to-7 correlations for PA were similar for the African-American and Caucasian girls. The year 2-to-7 correlation for the African-American boys was higher than for the Caucasian boys. The year 2-to-7 ICC reliability coefficients were moderate for V̇O2max for both genders and both ethnicities. The ICC for physical activity based on years 2–7 were not as strong as for V̇O2max and followed a similar trend to the Spearman ρ coefficients.
TABLE 3: Year-by-year Spearman correlations (r andP values), as well as first-to-last years correlations (Rho) and intraclass reliability coefficients (ICC) for aerobic power (mL·kg−1·min−1) and physical activity (PA) score presented by sex and ethnicity.
The subjects were divided into high, moderate and low fitness (V̇O2max) and physical activity levels by sex and ethnicity. The year-to-year κ values for stability of V̇O2max and PA are presented in Table 4 by gender and ethnicity. The κ values for V̇O2max indicated “substantial” (κ ∼ 0.50–0.615) year-to-year agreement on categorization for girls of both ethnicities. The Caucasian boys had greater agreement than the African-American boys, with the agreement declining to the “moderate” range during the final years. The overall κ (yr 1–7) for V̇O2max for both genders and ethnicities would be considered “low,” especially for the African-American boys. However, the 95% confidence intervals suggest that were significant. The κ values for physical activity categorization indicated “low to moderate” year-to-year agreement for both sexes and ethnicities. Agreements for years 2–7 were lower than the year-to-year. The overall, years 2–7, κ values for the girls of either ethnicity were quite low and not significant for the African-American girls (95% CI = −0.13 to 0.20), whereas the κ values for the boys were significant, but still low.
TABLE 4: Kappa statistics (κ) for the year-to-year stability of the youth to remain in the same group (low, moderate, and high) for aerobic power (V̇O2max) and physical activity levels over the 7-yr period; the year 1–7 κ with 95% confidence intervals is also presented.
The stability coefficients for tracking of V̇O2max and PA levels derived from the GEE are presented in Table 5. Tracking of aerobic power was highly significant and fairly similar between the sexes and ethnicities. The African-American youth did exhibit a slightly greater trend for tracking for both V̇O2max and PA than the Caucasian youth. The stability coefficients for tracking of PA level were also highly significant. Physical activity levels tracked better for boys than girls and slightly better for the African-Americans than Caucasians.
TABLE 5: Stability coefficients derived from the general estimating equation (GEE) analyses and 95% confidence intervals for tracking of aerobic power (V̇O2max) and physical activity over the 7-yr period, presented by sex and ethnicity.
DISCUSSION
The transition from elementary school to high school (typically occurring between ages 9 and 13 yr) is a volatile time. Physiology is changing dramatically, sexual maturation is occurring, and social and peer pressures increase. These changes typically are associated with a decline in physical activity levels and aerobic power (5,6,12,23). Our data would agree, as physical activity levels were reduced by about 48–53% for both the girls and boys, whereas aerobic power declined by approximately 18% for the girls and 7% for the boys. Our previous report shows an increase in sedentary activities during this time frame that could contribute to the reduction in moderate-to-vigorous PA levels and aerobic power (5). Furthermore, elementary school (kindergarten to grade 5) may present more opportunities for a wide variety of physical activities than high school (grades 9–12). For example, the elementary schools in our study had both recess and PE classes, whereas in our middle schools (grades 6–8) there was no recess and PE was limited to two to 3× wk−1. The high schools required only two semesters over 4 yr and the emphasis was on competition and high-level performance. Thus, physical activity opportunities within the school day diminished as the youth aged.
Bloom (4) suggested that Spearman correlations (ρ) ≥ 0.5 represent a stable characteristic if measurements are made at least 1 yr apart. For V̇O2max, our year-to-year correlations were well above 0.5. In addition, the κ scores suggest substantial stability within the groups from year to year (20). Thus, these analyses suggest substantial tracking of aerobic power. Although the year-to-year analyses suggested considerable stability, the year 1-to-7 ρ and κ coefficients suggest some change over the 7-yr period. The GEE stability coefficients were slightly above 0.5, suggesting only moderate stability and some change.
The results for physical activity are not as clear. Our year-to-year correlations for PA score vary widely, and the correlations (ρ and κ) from initial to final testing are low, well below the recommendation of Bloom (4). In addition, the κ scores suggest less stability within groups. However, the GEE stability coefficients for the boys were above 0.5, whereas the results of the girls were much lower. Pate et al. (21) noted that although absolute PA levels declined over 3 yr, some evidence of tracking was seen, as their Spearman correlations ranged from ∼ 0.57 to 0.61. Conversely, Janz et al. (10) noted only marginal correlations between PA levels measured over 4 yr (r ∼ 0.32–0.43). Our correlations from years 2-to-7 are somewhat lower than these other studies and could be related to the self-report instrument or greater heterogeneity of our statewide sample. Our results do agree with the findings of others, that as the interval of time increases between pre- and postmeasures, the tracking correlations decrease (10,14,28).
Although the values for V̇O2max declined as the youth aged, the statistical results for tracking of V̇O2max were stronger than for physical activity. Janz et al. (10) completed a 5-yr tracking study and also found that the correlations were stronger for V̇O2max than for physical activity. Twisk et al. (29) suggested that the strength of association is best quantified by looking at the confidence intervals (CI) for GEE analyses. As shown in Table 5, the limits of the CI are much closer for V̇O2max than for PA, indicating a stronger association. There are several reasons for these differences between the V̇O2max and PA results. First, V̇O2max has a genetic component that limits the ability to modify aerobic power (13), whereas physical activity is a behavior for which there is a greater variability of responses. Second, the measure of relative aerobic power includes weight (mL O2·kg−1body mass·min−1) and there is tracking of weight. We found a Spearman correlation of 0.767 (P = 0.0001) between the initial and final weight. Thus, higher weight youth tended to have higher weights as they aged and consequently had low relative aerobic power. Third, aerobic power was measured more empirically and with greater precision that physical activity. Finally, physical activity was obtained by a questionnaire and required recall of previous experiences, which may not be as accurate in younger participants, the early years of the study.
Ethnicity.
To our knowledge, no other study has attempted to evaluate tracking of V̇O2max and PA levels in African-American youth. The V̇O2max of all girls declined between years 1 and 7, but the African-American girls had a greater decline than the Caucasians. The decline was not related to a greater drop in PA scores for the African-American girls, as their final PA scores were higher than the Caucasians (Table 2). The African-American girls experienced a greater weight gain than their Caucasian counterparts, which may have contributed to their lower V̇O2max.
Aerobic power of African-American and Caucasian youth appear to track similarly, as the GEE stability coefficients were similar between ethnicities (Table 5). However, the Spearman correlations and κ statistics for the sex within ethnic groups were not as consistent, suggesting some differences. Thus, a closer look at group trends seems appropriate. Although V̇O2max declined with age, African-American and Caucasian girls tracked similarly, as indicated by similar Spearman correlations and stability coefficients. That is to say, girls retained their relative ranks within the group. Figure 1, however, shows that more African-American girls remained in the low-fit group over the 7-yr period than Caucasian girls (73 vs 58%, respectively) and more African-American girls migrated downward from the high-fit group to the low-fit group than the Caucasian girls (19% vs 11%, respectively). The change in aerobic power for the African-American girls was not related to PA levels, as those in the Low PA group tended to migrate upward into the moderate PA classification (Fig. 2). The earlier onset of puberty or greater body mass of the African-American girls may have contributed to the decline in aerobic power.
FIGURE 1: Fitness status at the seventh year of youth that were determined in year 1 to be in the lowest tertile (low V̇O2max; top graph) or in the highest tertile (high V̇O2max; bottom graph) presented by sex and ethnicity (AA, African-American; CA, Caucasian).
FIGURE 2: Physical activity levels at the seventh year of youth that were determined in year 1 to be in the lowest tertile (low PA; top graph) or in the highest tertile (high PA; bottom graph) presented by sex and ethnicity (AA, African-American; CA, Caucasian).
The results for aerobic power in the boys were different. Although the GEE stability coefficient was similar between ethnicities, the African-American boys had lower Spearman correlations and κ scores when comparing the first to last years. Figure 1 shows that more Caucasian than African-American boys tended to remain in their respective fitness groups (low or high) over the 7-yr period. In other words, the decline in V̇O2max for the Caucasian boys tended to occur in all three groups. In contrast, the African-American boys tended to maintain, or even increase, their V̇O2max over the same time interval, as more low-fit African-American boys moved to a higher group than Caucasians (∼60% vs 36%, respectively). There were more low SES African-American boys than Caucasian boys. Potentially, these lower SES African-American boys may have seen physical prowess as a way to improve their status (possibly for professional sports or university admissions). This is purely speculation and further study is needed for verification. As with the girls, these changes in V̇O2max appear not to be related to changes in PA levels. Because we are unaware of other reports examining ethnic differences in tracking of aerobic power, it is difficult to confirm our findings.
The tracking for physical activity level varies by ethnicity (Fig. 2). As with V̇O2max, physical activity levels declined over time, such that a PA score in the seventh year that indicated a high PA group (e.g., a boy with a score of 236) would actually be considered a score representative of the low PA group in initial years of the study, in which the range for low PA group was 177 to 303. It appears that more African-Americans than Caucasians maintained (or increased) their PA levels over they course of the study, as the proportion of the initial low PA group remaining in the low PA group at the end of the 7-yr period was less than the Caucasian youth (∼25% vs 44%, respectively). Conversely, more Caucasian youth remained in the low PA group over the same time period. With regard to those initially in the high PA group, there was a trend for all girls, regardless of ethnicity, and for Caucasian boys to migrate toward lower activity groupings as they aged. This may possibly be related to societal or pressures on the girls, or a tendency for the Caucasian boys to diversify their interests. However, the African-American boys who were initially active remained fairly active 7 yr later. Thus, tracking of high levels of PA appears to be strongest for African-American boys. Because the proportion of low SES was higher in the African-American boys, one could speculate that the higher tracking may be related to their continued interest in competitive sports, to enhance status, or self-esteem. However, there is no proof for this speculation.
The four methods to evaluate tracking show some disparity in their results. The GEE stability coefficients as determined by the regression coefficient (β) produced the highest tracking values for both V̇O2max and PA. This form of analysis takes into consideration the within-subject correlation and may be more robust in that “the point estimates of the model parameters are usually similar” (29). However, this approach may work better with irregularly spaced time periods, rather than measurements taken 1 yr apart (29). The Spearman ρ statistic has been used for quite some time to examine stability of human characteristics (4). In addition, the reliability ICC has become an accepted method to evaluate tracking. Our results using these two techniques (ρ and ICC) suggest that tracking for V̇O2max is considerably higher than for PA. Similarly, the κ statistic results used for categorical tracking also suggest that tracking for V̇O2max is considerably higher than for PA. The Spearman ρ, ICC, and κ methods appear to fit our data better than the GEE method, as Figure 1 shows that a major proportion of the fitness groups remained in their respective grouping at the end of the 7 yr, whereas Figure 2 shows a lower proportion of the initial PA grouping remained in the same group 7 yr later. When all these methods are taken together, we suggest that the Spearman ρ, ICC, and κ statistic may be more appropriate for equally spaced data, whereas the GEE β coefficient may be better suited for irregularly spaced data.
Tracking implies that the active remain active and the inactive remain inactive, the fit remain fit and/or the reciprocal. Strong tracking suggests that change may not likely occur, whereas moderate tracking indicates that change is possible. From a health perspective, high tracking of V̇O2max or PA levels would suggest that efforts to induce change should occur early, before tracking is ingrained, as later efforts may not produce successful results. However, moderate-to-low tracking could be a beneficial characteristic, as it suggests that individuals with the lowest V̇O2max, or inactive individuals, have the potential to increase these two characteristics, a desirable outcome. Moderate tracking also implies that individuals with high V̇O2max, or who are highly active, could lower their aerobic power or become less active, a less than desirable outcome. The optimal characteristics would be that those individuals with low aerobic power, or those who are the least active, could improve whereas those with high levels would maintain. Our V̇O2max data (Fig. 1) suggest a less than desirable trend, as there was a greater tendency for the low-fit group, particularly African-American girls, to remain low fit than for the high-fit group to remain high fit. However, our PA data (Fig. 2) suggest the possibility of a desirable trend, as there was greater movement from the least active African-Americans to a more active group, whereas the most active group remains more stable. Thus, a potential attainable goal would be to motivate the least active individuals, particularly African-American girls to improve the levels of PA, while maintaining the activity levels in the most active group.
Limitations.
There are some limitations to the interpretation of these results. First, although the initial enrollment of African-Americans was fairly significant (N = 428), we had complete data over the 7-yr period on only 153 (75 girls and 78 boys). Thus, the generalizability of our results may be somewhat compromised and some researchers might consider the results as “exploratory.” Other researchers, however, have reported on similar or smaller sample sizes than our African-American groups and found significant results (10,21,24). We do not feel that attrition biased our results, as the African-American subjects had V̇O2max and PA scores similar to the overall African-American sample obtained from across the state and those that were lost at follow-up (Table 1). Second, physical activity was obtained by a questionnaire that evaluated habitual activity. Although there is inherent variability in the ability to recall activities, this approach as been used successfully in other studies (10,12,21). Third, our analyses did not control for socioeconomic status. Sallis et al. (24) have shown that low SES modifies opportunities for physical activities. In defense, both our initial sample and the subset included a broad spectrum of SES. Although we retained more African-American low SES boys in our sample, the SES levels of the rest of the sample lost at follow-up and subset examined in this paper were similar (Table 2). Therefore, SES should have had minimal impact on the results. Fourth, we did not adjust any of the results for pubertal status because we did not have complete developmental stage data on our sample. Puberty, which causes body composition and psychological changes, could help us understand why the differences in tracking, because it is known that African-American youth tend to develop earlier than Caucasian youth. Finally, V̇O2max was estimated from a submaximal cycle ergometer test, which has some inherent error (18). Although estimating V̇O2max, rather than a direct measurement, reduces accuracy and precision, the test-retest reliability coefficient is high, >0.90 (17). Thus, the method provides a highly reproducible estimate of V̇O2max.
Our study does provide some important information. We have shown that although aerobic power and physical activity levels decline from childhood through adolescence, aerobic power tracks better than physical activity levels. African-American girls who are low fit (low V̇O2max) have a greater tendency to remain low fit as they age. Initially, low-fit African-American boys tend to maintain their V̇O2max levels as they age, and because V̇O2max generally declines with age, their rank order increases in the population, moving them to a higher fitness group. With regard to physical activity levels, there appears to be similar stability for Caucasian and African-American girls, but African-American boys have a greater tendency to maintain, or possibly increase, activity levels as they age. In addition, highly active African-American boys tend to remain highly active. Thus, in general, it appears that inactive and unfit youth become more inactive and unfit teenagers, particularly African-American girls. The positive finding of this study is that tracking is only moderate, suggesting that change is possible. Programs that are gender and ethnic sensitive should be developed and tested to determine whether these could increase activity levels within each subpopulation. We can intervene to attempt to break the cycle of tracking, particularly in low fit and inactive youth and, at the same time, we can attempt to maintain tracking of the high fit and active youth.
The authors would like to acknowledge the assistance of Drs. Tom Baranowski, Robert Malina, and Hans Kemper in the preparation of this manuscript.
This study was supported by grant no. NR01837 from the National Institute of Nursing Research of NIH.
REFERENCES
1. Aaron, D., A. M. Kriska, S. R. Dearwater, et al. The epidemiology of leisure physical activity in an adolescent population. Med. Sci. Sports Exerc. 25: 847–853, 1993.
2. Ainsworth, B. E., W. L. Haskell, M. C. Whitt, et al. Conmpendium of physical activities: an update of activity codes and MET intensities. Med. Sci. Sports Exerc. 32: S498–S504, 2000.
3. Berkey, C. S., H. R. H. Rockett, A. E. Field, et al. Activity, dietary intake, and weight changes in a longitudinal study of preadolescent boys and girls. Pediatrics 105: E561–E569, 2000.
4. Bloom, B. Stability and Change in Human Characteristics. New York: Wiley & Sons, 1964, p. 3.
5. Bradley, C. B., R. G. Mcmurray, J. S. Harrell, S. Deng. Changes in common activities of 3rd through 10th graders: the CHIC study. Med. Sci. Sports Exerc. 32: 2071–2078, 2000.
6. Caspersen, C. J., M. A. Pereira, K. M. Curran. Changes in physical activity patterns in the United States by sex and cross-sectional age. Med. Sci. Sports Exerc. 32: 1601–1609, 2000.
7. Gilmer, M. J., B. J. Speck, C. B. Bradley, J. S. Harrell, M. Belyea. The
Youth Health Survey: reliability and validity of an instrument for assessing cardiovascular health habits in
adolescents. J. School Health 66: 106–111, 1996.
8. Gordon Larson, P., R. G. Mcmurray, B. M. Popkin. Adolescent physical activity and inactivity vary by ethnicity: the National Longitudinal Study of Adolescent Health. J. Pediatr. 135: 301–306, 1999.
9. Harrell, J. S., R. G. Mcmurray, S. A. Gansky, S. I. Bangdiwala, A. C. Frauman, C. B. Bradley. Public health versus a risk-based intervention to improve cardiovascular health in elementary school children: the Cardiovascular Health in Children study. Am. J. Pub. Health 89: 1529–1536, 1999.
10. Janz, K. F., J. D. Dawson, L. T. Mahoney. Tracking physical fitness and physical activity from childhood to adolescence: the Muscatine Study. Med. Sci. Sports Exerc. 32: 1250–1257, 2000.
11. Kemper, H. C. G., W. Devante, W. Van Mechelen, J. W. R. Twisk. Adolescent motor skill performance: is physical activity in adolescence related to adult physical fitness? Am. J. Hum. Biol. 13: 180–189, 2001.
12. Kimm, S. Y., N. W. Glynn, A. M. Kriska, et al. Longitudinal changes in physical activity in a biracial cohort during adolescence. Med. Sci. Sports Exerc. 32: 1445–1454, 2000.
13. Klissouras, V. Hertability of adaptive variation. J. Appl. Physiol. 31: 338–344, 1971.
14. Malina, R. Tracking of physical activity across the lifespan. Research Digest, President’s Council on Physical Fitness and Sports, Series 3: 14, 2001.
15. Maia, J. A. R., J. Lefevre, A. Claessens, R. Renson, B. Vanreusel, G. Beunen. Tracking of physical fitness during adolescence: a panel study in boys. Med. Sci. Sports Exerc. 33: 765–771, 2001.
16. Maia, J. A. R., J. Lefevre, G. Beunen. Stability of physical fitness: a study of Belgium males followed longitudinally from 12 to 30 years. Med. Sci. Sports Exerc. 30: S305, 1998.
17. Maynard, E. J., W. E. Coonan, A. Worsley, T. Dwyer, P. A. Baghurst. The development of the lifestyle education program in Australia. In: Cardiovascular Risk Factors in Childhood: Epidemiology and Prevention, B. Hetzel, and G. S. Berenson (Eds.). New York: Elsevier, 1987, pp. 123–148.
18. Mcmurray, R. G., W. K. Guion, B. E. Ainsworth, J. S. Harrell. Predicting aerobic power in children. J. Sports Med. Phys. Fitness 38: 227–233, 1998.
19. Mcmurray, R. G., J. S. Harrell, C. B. Bradley, S. Deng, S. I. Bangdiwala. Predicted maximal aerobic power in
youth is related to age,
gender and ethnicity. Med. Sci. Sports Exerc. 34: 145–151, 2002.
20. Munoz, S. R., S. I. Bangdiwala. Interpretation of Kappa, and B-statistics measures of agreement. J. Appl. Stat. 24: 105–111, 1997.
21. Pate, R. R., T. Baranowski, M. Dowda, S. G. Trost. Tracking of physical activity in young children. Med. Sci. Sports Exerc. 28: 92–96, 1996.
22. Ross, J. G., G. G. Gilbert. The National Children and
Youth Fitness Study: a summary of findings. J. Phys. Educ. Rec. Dance 56: 45–50, 1985.
23. Sallis, J. F. Age-related decline in physical activity: a synthesis of human and animal studies. Med. Sci. Sports Exerc. 32: 1598–1600, 2000.
24. Sallis, J. F., L. Greenlee, T. L. Mckenzie, et al. Changes and tracking of physical activity across seven years in Mexican-American and European-American mothers. Women & Health 34: 1–14, 2001.
25. Shrout, P. E., J. L. Fleiss. Intraclass correlations: uses in assessing rater reliability. Psychol. Bull. 86: 420–428, 1979.
26. Shvartz, E., R. C. Reibold. Aerobic fitness norms for males and females aged 6 to 75 years: a review. Aviat. Space Environ. Med. 61: 3–11, 1990.
27. Sunnegardt, J., L. E. Bratteby, S. Sjolin. Physical activity and sports involvement in 8- and 13-year-old children in Sweden. Acta Paediatr. Scand. 74: 904–912, 1985.
28. Telama, R., E. Leskinen, X. Yang. Stability of habitual physical activity and sport participation: a longitudinal tracking study. Scand. J. Med. Sci. Sports 6: 371–378, 1996.
29. Twisk, J. W. R., H. C. G. Kemper, W. van Mechelen, G. B. Post. Tracking of risk factors for coronary heart disease over a 14-year period: a comparison between lifestyle and biologic risk factors with data from the Amsterdam Growth and Health Study. Am. J. Epidemiol. 145: 888–898, 1997.
30. Trowbridge, C. A., B. A. Gower, T. R. Nagy, G. R. Hunter, M. S. Treuth, M. I. Goran. Maximal aerobic capacity in African American and Caucasian prepubertal children. Am. J. Physiol. 273: E809–E814, 1997.
