Study 1
A 2004 study by Elliot and colleagues included 18 public high schools from northwest Oregon and southwest Washington who participated in a prospective randomized trial of the Athletes Targeting Healthy Exercise & Nutrition Alternatives (ATHENA) intervention. Schools were matched in pairs based on size, socioeconomic status, and student demographics, and then were randomly assigned to either the treatment or control group. Schools that were randomized into the treatment group were offered the chance to allow all female student–athletes in sports, dance, and cheerleading teams to participate in the ATHENA program. At least one team per school agreed to participate. For every treatment team that agreed to participate in the ATHENA program, a similar team from the matching control school agreed to participate in an alternative condition. The control teams were offered preprinted pamphlets concerning disordered eating, drug use, and sports nutrition. Participating schools came from various locations to minimize cross-contamination.
The study examined the immediate posttest outcomes from the year 1 cohort. The intervention group consisted of nine schools, with 457 students surveyed in the preseason and 337 students surveyed in the postseason (74 percent retained). The average age of the female students in the intervention group was 15.4 years, and 93.6 percent were white. The control group consisted of nine matched schools, with 471 students surveyed in the preseason and 331 students surveyed in the postseason (70 percent retained). The average age in the control group was 15.3 years, and 91.4 percent were white. Before the ATHENA intervention, the treatment and control groups reported similar demographics, drug use indices, knowledge scores, and self-assessed behaviors and attitudes.
Immediate outcomes were assessed with pre–sports and post–sports season self-report questionnaires. The pretest was given before participants began the ATHENA program; the posttest measures were gathered within 2 weeks following the end of the sport season. The questionnaires assessed knowledge, potential influences on drug use and disordered eating (including those of peers, coaches, the media, and society), and participant characteristics. The survey questions were based on items from earlier research and other surveys concerning disordered eating, depression, self-image, and self-esteem. Behavioral intent was assessed by asking about future actions using a seven-point Likert scale, with responses ranging from “strongly disagree” to “strongly agree.” Substance abuse questions used a format similar to Monitoring the Future and the Youth Risk Behavior Surveillance to assess supplement use and disordered eating behaviors.
To determine baseline equivalence and intervention effects, the study used an analysis of covariance–based approach within the generalized estimating equations random-effects model framework, which extends generalized linear models to deal with time dependence and hierarchical structure. The incidence of new users (i.e., athletes who began a behavior after the baseline assessments) was determined and conditions were compared using a chi-squared test analysis. An analogous format was used to examine participants who stopped a behavior. No subgroup analysis was conducted.
Study 2
The 2008 study by Elliot and colleagues was a follow-up to Study 1 (Elliot et al. 2004) that examined the short-term effectiveness of ATHENA. The 2008 study analyzed the results from a follow-up survey that was sent to all treatment and control participants who had completed high school and were 18 years and older 1 year after the ATHENA intervention’s last year. A one-page anonymous survey was sent, along with a self-addressed, stamped envelope, to participants’ original addresses. The mailed questionnaire was brief and was limited to demographics, drug use, eating behaviors, and selected attitudes.
The follow-up surveys were mailed to 368 students in the intervention group; 203 (55 percent) returned responses. The surveys were also mailed to 389 students in the control group, and 197 (51 percent) returned responses. For treatment group participants who returned the follow-up survey, 90 percent were in college or trade school, and 51 percent had graduated from high school a year before responding to the survey, while 49 percent had graduated 2 or 3 years before. For control group participants, 90 percent were in college or trade school, and about 45 percent had graduated from high school a year before responding to the survey, while 55 percent had graduated 2 or 3 years earlier.
Alcohol use, tobacco use, other drug use, and disordered eating behaviors were indexed using the original standardized format. For less-frequent behaviors, such as smoking and marijuana use, the study compared use in the last year and lifetime use. For more prevalent behaviors, such as alcohol use, use in the past 3 months and in the last year were indexed and compared. The questionnaire also presented respondents with a spectrum of female physiques and asked them to identify the healthiest and most attractive representation.
Baseline equivalence and shorter-term intervention effects were examined using an analysis of covariance–based approach within the generalized estimating equations random-effects model framework. Students were no longer clustered on teams and in schools. The association between participants’ study condition and their long-term drug and alcohol use was examined with logistic regression using years since graduation as a covariate in the analysis. Body image responses were assessed with t–tests. No subgroup analysis was conducted.