describes characteristics of the sample by sex, grade, and ethnicity. The sample included a similar number of female and male subjects. The ethnic distribution of the children was 59% non-Hispanic white, 27% Hispanic, 7% African American and 6% Asian/Pacific Islander. To investigate whether the Fitnessgram participants were demographically representative of the district, the distribution of the sample by ethnicity was compared with the ethnic distribution of the district. The distributions did not differ significantly by ethnicity (P
> 0.15), but did vary by free and reduced price meal eligibility (P
= .03), suggesting that the Fitnessgram participants were ethnically representative of the students comprising the school district but slightly better off, economically, than the school district as a whole (). The frequency of parental completion of college was 61.5% and only 22.8% of the children were eligible for free or reduced price school lunches, indicating a school district with higher socioeconomic status than the average California public school district. The average California school district in 2002–2003 had 48.3% of students eligible for free or reduced price school lunches (obtained from http://www.cde.ca.gov/ds/sh/cw/filesafdc.asp
illustrates the mile time by healthy fitness zone standards, as established by the state of California (8
). Physical education instructors were instructed to stop the Fitnessgram mile run/walk test at 15 minutes; some physical education instructors permitted students completing the final quarter mile to complete it in 16 minutes. Observed means are therefore underestimates, given that some students who were unable to complete a mile because of the time constraint would have completed it in more time if permitted. Sixty-five percent of students had a fitness level below recommended age-specific, sex-specific standards for mile time performance. Additionally, 64% of study participants had mile times that were slower than the norms recommended by the state of California. There was a similar percentage of girls and boys classified as being in the healthy fitness zone (P
=0.48). also depicts the relationship between ethnicity and achievement of California age-specific mile time performance. African American students were less likely to achieve California fitness standards than Asian American and non-Hispanic white students (ORAsians
= .49, P
= .04; ORWhites
= .58, P
Fitness and BMI-for-age classifications by sex and ethnicity.
Obesity status and mean BMI-for-age percentiles (converted from z-scores for ease of interpretation) are also included in . The combined prevalence of overweight and obese (BMI≥85th percentile) was 31.8% for males and 27.7% for females (). Mean BMI-for-age percentiles varied by ethnicity, with 16.3% of Asians classified as either at risk for overweight or obesity, 22.1% of non-Hispanic whites, 40.6% of African Americans and 47.7% of Hispanics. The percentages of students classified as overweight or obese was greater among African Americans and Hispanics than among Asians or non-Hispanic whites (all comparisons P <0.003, after Bonferroni correction). Additionally, Asian/Pacific Islanders had slightly lower BMI-for-age percentiles than non-Hispanic whites (F(1,11) = 10.65, P = .04). Hispanics and African Americans did not differ significantly from each other, nor did boys’ BMI-for-age differ on average from that of girls.
With respect to age-specific fitness standards, students who failed to run the mile in the appropriate time interval established as appropriate for each age and sex scored significantly lower on the CAT6 and CST math, reading and language California standards tests compared with those students who fell in the healthy fitness zone (). Tests for linear trends revealed that decreasing quintiles of aerobic fitness scored progressively lower on CAT6 math and reading (linear trend, Pmath
< .0001; Preading
= .001) () and on CST math and language tests (linear trend, Pmath
< .0001; Planguage
< .0001) (; available at www.jpeds.com
Mean¶ performance on standardized math, reading and language tests, by aerobic fitness classification§ and by BMI-for-age obesity status.‡
Figure 1 CAT6 mean scores, adjusted by the covariates sex, free and reduced price lunch eligibility, and ethnicity. A, Percentiles in CAT6 score (math) by quintiles of minutes to complete mile run; * 3rd quintile of mile time differed on 2002 CAT6 math score from (more ...)
Figure 2 CST mean scores, adjusted by the covariates sex, free and reduced price lunch eligibility, and ethnicity. A, Percentiles in CST score (math) by quintiles of minutes to complete mile run, *3rd quintile of mile time differed on 2002 CST math score from (more ...)
depicts test scores by BMI percentiles; as observed for mile run/walk time, those who exceeded both the 85th and 95th percentile for BMI-for-age scored significantly lower on the CAT6 and CST math, reading and language tests than those in the recommended range for BMI. Tests for linear trend showed that increasing quintiles of BMI-for-age percentile scores scored progressively lower on both CAT6 math and reading (linear trend, Pmath = .007; Preading = .028) () as well as CST math and language tests (linear trend, Pmath = .013; Planguage = .073) ().
Sequential hierarchical linear regression models (Stata xtreg procedure) were used to regress CAT6 and CST test score performance measures onto the following predictors: 1) null model, 2) age, ethnicity, sex, and eligibility for free/reduced price school lunches, 3) the foregoing covariates and BMI-for-age z-scores, and 4) the foregoing covariates/predictors and mile run/walk time. Mile run/walk time was a significant predictor of standardized CAT6 2002 math test score performance such that the math score dropped 1.9 points (out of a possible 99) for every additional minute required to complete the one mile run/walk (b = −1.94, 95% CI = −2.37, −1.53) even when age, free or reduced price lunch status, sex, and ethnicity and BMI-for-age were included as covariates. Adding the demographic covariates to the null model reduced the intraclass correlation from .09 to .01 and explained 15.9% of the null model variance. Adding BMI-for-age to the demographic variables-augmented model decreased the null model variance only an additional 0.4%, although this change was still statistically significant (model difference likelihood ratio chi square (1) = 8.6, P = .003). The full model, including all of the foregoing covariates/predictors (including BMI-for-age) but also including a measure of the student’s performance on the 1-mile fitness test decreased the null model variance an additional 3.5% (model difference likelihood ratio chi square(1) = 81.6, P < .0001). With the inclusion of the fitness measure, the student’s BMI-for-age was no longer a significant contributor to the CAT6 2002 math test score. All major ethnic groups differed significantly from Whites in the full model, with Asian math scores higher than Whites’ scores (bAsian = 4.6, 95% CI = 1.44, 7.85), and Hispanic and African American math scores lower than Whites’ scores, respectively (bHispanic = −11.44; 95% CI = −13.74, −9.14; bAfrican American = −16.33, 95% CI = −19.70, −12.96).
The standardized CAT6 2002 reading test score dropped 1.1 points for every additional minute required to complete the one mile run/walk (b = −1.13, 95% CI = −1.56, −0.70) even when age, sex, ethnicity, free and reduced price lunch status, and BMI-for-age were included as covariates. Adding the demographic covariates to the null model reduced the intraclass correlation from .09 to .01 and explained 15.6% of the null model variance. Adding BMI-for-age to the demographic variables-augmented model decreased the null model variance only an additional 0.7%, although this change was still statistically significant (model difference likelihood ratio chi square (1) = 6.23, P = .013). The full model, including all of the foregoing covariates/predictors, including BMI-for-age, but also including a measure of the student’s performance on the 1-mile fitness test decreased the null model variance an additional 1.3% (model difference likelihood ratio chi square(1) = 26.2, p < .0001). With the inclusion of the fitness measure, the student’s BMI-for-age was no longer a significant contributor to the CAT6 2002 reading test score. Asian ethnicity generally explained no additional variance in CAT6 reading test scores relative to Non-Hispanic whites, which was the referent ethnic group, but Hispanic and African American ethnicity did explain some additional variation (11.2 point drop for Hispanics (bHispanic = −11.23; 95% CI = −13.57, −8.89); 14.5 point drop for African Americans (bAfrican American = −14.53, 95% CI = −17.96, −11.12).
Similar findings were noted for CST math and language test performance (data not shown). These analyses also confirmed that the pattern of ethnic differences in standardized test scores (Asians and Non-Hispanic Whites > African Americans and Hispanics) was consistent with the pattern of ethnic differences in percent achieving recommended levels of BMI and aerobic fitness.