Study subjects for this investigation come from the Epidemiology of Physical Activity from Adolescence to Adulthood study, a 20 year longitudinal cohort that has been followed from adolescence to adulthood and has been assessed during three separate cycles (phases). The original subjects, aged 12 - 16 years [n = 1245 adolescents (89% of the total student population)] were recruited in 1989 when they were enrolled in junior high school at a single school district in Pittsburgh, PA and followed for a period of four years (Phase I). The cohort consisted of similar numbers of male (n = 641) and female (n = 604) adolescents; and the racial composition was 73% white, 24% African American, and 3% Hispanic or Asian [15-17]. In 1999, subjects were re-contacted (aged 22 - 25 years of age) to participate in a follow-up study to examine changes in physical activity from adolescence to young adulthood (Phase II). A total of 827 (66%) completed an interviewer-administered, follow-up questionnaire which included information about health behaviors (i.e., smoking, drug and alcohol use, sedentary behavior), physical activity. Each phase of the current study was approved by the University of Pittsburgh Institutional Review Board and written informed consent was obtained from all participants and/or their parents prior to participation in any part of the study.

Physical activity (PA) was assessed using the Modifiable Activity Questionnaire for Adolescents (MAQ-A) [18] and the Modifiable Activity Questionnaire (MAQ). The MAQ-A is an adolescent-specific questionnaire based on the Modifiable Activity Questionnaire (MAQ) for adults. The MAQ-A has been shown to provide reproducible and valid estimates of past year physical activity in adolescents [15,19]. For each of the four years during Phase I, competitive activities as well as leisure time physical activity data were collected by administering the MAQ-A during the spring of each year. Trained research assistants supervised the students as they completed the questionnaire during their regular physical education class. Students were asked to report the frequency and duration of all activities they participated in at least ten times over the past year. The questionnaire inquired about time spent engaged in competitive activities and leisure time PA. The estimated number of hours spent on average for each activity per day was calculated and the average hours per day of all activities was summed to give an estimate of the total physical activity in hours per week averaged over the past year.

During the 1999 follow-up (Phase II), trained interviewers administered a past year Modifiable Activity Questionnaire (MAQ) [20,21]. The MAQ, an interviewer-administered questionnaire for adults, assesses both leisure and occupational activities similar to the MAQ-A completed in the AIC study. The 1999 version asked participants to indicate activities that they had participated in at least ten times over the past year. For each activity, the months of participation was indicated as well as the average days per week and average minutes per day of participation. To assess occupational (non-leisure) activities, participants were also asked to list all jobs held for over one month during the past year and answered questions related to transporttation to and from work, days per week and hours per day of work, time spent sitting, and intensity of non-sitting work related activities.

Health risk behaviors were assessed during years 1, 2, and 4 of Phase I using the Center for Disease Control and Prevention’s Youth Risk Behavior Survey [22] which has been tested for reliability and validity [23]. The health risk behavior questions used in this analysis measured tobacco, alcohol, and binge drinking in the past 30 days. Identical questions were used during the Phase II follow-up phone interview.

During the Phase I, height (cm) and weight (kg) were measured annually with a standard balance scale and used to calculate body mass index (BMI; kg/m²). Shoes were removed prior to all measurements of height and weight. During the Phase II follow-up, BMI was calculated from self-reported height and weight on the questionnaire.

Only participants that had BMI and complete physical activity data at all time-points during Phase I and Phase II (years 1990-1993 and 1999/2000) were included in the analyses. All continuous data were assessed for normality. Normally distributed data are reported as mean (SD), non-normal variables as median (Interquartile range). Categorical data are presented as percentages. Descriptive statistics were calculated for the total cohort and sex-specific grouping. Comparisons between sexes of continuous variables were assessed using a twosample t-test or the Wilcoxon test. Categorical variables were assessed using the Pearson chi-square test. Evaluation of the change in HRB from adolescence to adulthood was performed using a Mc-Nemar’s test.

A series of gender-specific logistic regression models were utilized to evaluate the independent impact of PA, BMI, and adolescent health risk behavior (HRB) on adult HRB. Independent variables included in the regression analyses were treated categorically. Gender-specific physical activity levels were divided into “high” and “low” groups using a median split. BMI was categorized into “normal” (<25.0 kg/m²) and “overweight/obese” (≥25.0 kg/m²) according to NHLBI BMI guidelines [24]. Adolescent health risk behaviors were coded as “never” and “ever” for each heath risk behavior of interest (smoking, drinking, and binge drinking). The outcome of interest, adult HRB, was also coded as “never” and “ever” for each heath risk behavior of interest (smoking, drinking, and binge drinking) during the four year period. Separate univariate models for each HRB of smoking, drinking, and binge drinking were constructed for PA alone, BMI alone, and adolescent health risk behavior alone. The next series of models were conducted using a combination of variables. Model 4 included PA and BMI. Model 5 included PA and adolescent health risk behavior and Model 6 included BMI and adolescent health risk behavior. The final model included PA, BMI, and adolescent HRB. We tested for statistical interactions between physical activity and adolescent health risk behavior by adding a two-way interaction term to the model. Assessment of these interactions was determined by examining the individual p-values of the interaction coefficients and by assessing the impact of the addition of this term to the overall fit of the model. Statistical analyses were performed using Statistical Analysis Software, version 8.2 (SAS Institute Inc., Cary, North Carolina). Statistical significance was considered as a p-value <0.05.

A total of 1245 individuals participated in Phase I, of which 828 participants (67% of original Phase Icohort) completed a follow-up questionnaire, in adulthood, as part of the Phase II. Complete data were available for 536 participants (Figure 1).

Descriptive statistics for the 536 subjects are provided from Phase I (Table 1(a)) and Phase II (Table 1(b)). The mean age of the cohort at the Phase II was 24.8 (±1.1) years with an even split between males and females. Overall, the sample was predominately Caucasian with approximately 74.3% reporting working full time and 62.4% reporting a degree past high school. The average adolescent BMI was 22.5 kg/m² and increased to 25.4 kg/m² during adulthood. Physical activity for the total cohort decreased from a median of 12.9 hours per week during adolescence to 5.6 hours per week during young adulthood (Tables 1(a)-(b)).